Important Dates


Paper Submission Deadline: 14th Feb 2021
28th Feb 2021
07th Mar 2021 (Final deadline)
Notification of acceptance: 10th May 2021
Final Manuscript Due: 25th May - 15th June 2021






















Contributed Papers


Title Authors Abstract Session title Room Time
Enhanced Precoding Aided Generalized Spatial Modulation for Massive MIMO Systems Kunnathully Sadanandan Sanila and Neelakandan R (IIT Goa, India) Receive spatial modulation (RSM) is one of the most promising paradigms that significantly reduces the receiver's computational complexity. However, to assure the linear precoding operation at the transmitter side, RSM systems have to be under-determined. We propose a transmission scheme that divides antennas at the transmitter into G t transmit antenna groups (TAGs) and antennas at the receiver into G r receive antenna groups (RAGs) for exploiting the SM concept at the transceiver ends. Additionally, we extend the notion of generalized spatial modulation (GSM) to a new precoding-aided massive multiple-input multiple-output (mMIMO) system and formulate the structure, particularly in an activated antenna group at the transmitter and receiver. We refer to it as an enhanced receive GSM (ERGSM) system. The antenna grouping makes the proposed GRSM based scheme suitable for both the under-determined and over-determined massive MIMO architectures according to the distribution of the number of TAGs and RAGs and thus increases the resilience of the system. We project a low complexity sub-optimal detection algorithm for the proposed scheme. Further, we computed the complex calculations required for the system and compared them to the other conventional techniques. Also, we present numerical results to substantiate our ideas. 5G Communications Auditorium 7/29/21 16:00
K-Preferential Slotted ALOHA for Ultra Reliable Low Latency Communications Satyam Agarwal and Shubham Pandey (Indian Institute of Technology Ropar, India) Ultra-reliable low-latency communication (URLLC), one of the key component of 5G, provides a set of features required to support mission-critical applications. Slotted ALOHA is one of the most popular mechanism to share a channel among multiple users. However, slotted ALOHA cannot meet high reliability requirements of URLLC when large number of users are present in the network. In this paper, we propose a preferential medium access control scheme to match the high reliability requirement of URLLC. It is ensured by dedicating every K-th slot exclusively for URLLC transmission. We analytically obtain the packet delay distribution and reliability of both the URLLC and regular packets. An optimization problem is framed to maximize the reliability of regular packets subject to meeting the URLLC reliability constraints. Extensive simulations indicate that our proposed K-preferential S-ALOHA protocol can meet the URLLC requirements even when the network traffic is high. 5G Communications Auditorium 7/29/21 16:20
Uplink Channel Impulse Response Based Secondary Carrier Prediction Prayag Gowgi and Vijaya Parampalli Yajnanarayana (Ericsson Research, India) A typical handover problem requires sequence of complex signaling between a UE, the serving, and target base station. In many handover problems the down link based measurements are transferred from a user equipment to a serving base station and the decision on handover is made on these measurements. These measurements together with the signaling between the user equipment and the serving base station is computationally expensive and can potentially drain user equipment battery. Coupled with this, the future networks are densely deployed with multiple frequency layers, rendering current handover mechanisms sub-optimal, necessitating newer methods that can improve energy efficiency. In this study, we will investigate a ML based approach towards secondary carrier prediction for inter-frequency handover using the up-link reference signals. 5G Communications Auditorium 7/29/21 16:40
Phase Calibration of Multiple Software Defined Radio Transmitters for Beamforming in 5G Communication Dusari Nageswara Rao (Indian Institute of Technology, Roorkee, India); Meenakshi Rawat (IIT Roorkee, India) Beamforming is the key technique used in 5G communication systems for transmitting/receiving signals only in a particular direction. An accurate phase is needed to apply to the beamforming antenna array to steer the beam in a particular direction. Generally, multiple software-defined radios (SDR) are used for flexible beamforming. Whereas these multiple SDRs contain phase differences in transmitting paths due to nonlinearities in their components and the use of an individual clock and local oscillators (LO). Therefore, this paper presents the methodology to calibrate the phase differences in different transmitting paths of SDR before applying signals to the antenna elements for beamforming. This paper presents the methodology to estimate the phase offset using the cross-covariance method. A method is presented to synchronize multiple SDRs accurately. As a proof of concept, the SDR setup is built with the analog transceiver AD9371 from Analog Devices and ZC706 FPGA board from Xilinx. The measurement results with phase compensation after synchronization achieves an NMSE of around -35 dB between the signals of different transmitter paths. A 1×4 antenna array operating at 2.4 GHz has been designed in simulation, and the main beam is achieved in the desired direction after phase compensation. 5G Communications Auditorium 7/29/21 17:00
Improving the Throughput of a Cellular Network Using Machine Learning - A Case Study of LTE Prasad Tukaram Gaikwad (IITH, India); SaiDhiraj Amuru and Kiran Kuchi (IIT Hyderabad, India) Long Term Evolution (LTE) focused on providing high data rates at low latency when compared to previous-generation technologies. The recent research and development in machine learning for wireless communication networks focus on making these networks more efficient, intelligent, and optimal. We propose a machine learning algorithm to improve the performance of LTE in a real-time deployments. Specifically, we focus on the case of single-user multiple-input multiple-output transmission mode (TM4 as known in LTE). The channel quality feedback from user to the base stations plays a crucial role to ensure successful communication with low error rate in this transmission mode. The feedback from the user includes precoding matrix indicator (PMI), rank indicator apart from the channel quality feedback. However, in practical systems, as the base station must support several users, there is a delay expected from the time a user sends feedback until the time it is scheduled. This time lag can cause significant performance degradation depending on the channel conditions and also in cases when the user is mobile. Hence, to eliminate this adverse impact, we present a machine learning model that predict future channels and the feedback from the user is calculated based on these predictions. Via several numerical simulations, we show the effectiveness of the proposed algorithms under a variety of scenarios. Without loss of generality, the same work can be applied in the context of 5G NR. LTE is used only as a case study due to its vast prevalence and deployments even as of today. 5G New Radio Auditorium 7/29/21 14:30
Enhanced Transport Block Processing for 5G NR PUSCH Coverage Enhancement Koteswara Rao G and Subhash Kumawat (Indian Institute of Technology Hyderabad, India); SaiDhiraj Amuru and Kiran Kuchi (IIT Hyderabad, India) In 5G-New Radio (5G-NR), the available transmission power at the user is limited, and thus, the uplink transmissions define the coverage of 5G-NR systems. In this paper, we propose various solutions to enhance the uplink coverage of 5G-NR. Each payload of the user in 5G-NR is appended with a cyclic redundancy check (CRC) before channel encoding and transmission. We identify that the overhead with the CRC attached payload is as high as 40% in the cell-edge transmissions. To address this issue, we propose a mechanism to transmit the data over multiple slots with a single CRC attached to the payload. This method significantly reduces the CRC overhead and improves the cell-coverage. Through link-level evaluations, we show that this proposed method achieves 450 meters improvement in the cell-coverage for the VoIP scenario. Further, we identify the under-utilization of resources in the special slots. To reduce the reference signal overhead in the special slot transmissions, we have proposed allocating the resources by jointly considering the available symbols in special and uplink slots. Through link-level evaluations, we show that this proposed scheme achieves an improvement of close to 2 dB in the link performance and 400 meters improvement in the cell-coverage of the eMBB scenario. 5G New Radio Auditorium 7/29/21 14:50
Downlink Resource Allocation for 5G-NR Massive MIMO Systems Pavan Reddy M (Indian Institute of Technology Hyderabad, India); Mounika Reddy (IIT Hyderabad, India); Abhinav Kumar (Indian Institute of Technology Hyderabad, India); Kiran Kuchi (IIT Hyderabad, India) The gNodeB (gNB) in 5G-New Radio (5G-NR) is capable of beamforming and spatial multiplexing the users to achieve a multi-fold increase in the network capacity. With multiple active beams and the possibility of varying payload sizes, the resource allocation algorithm should optimally utilize the resources in time, frequency, and space. Otherwise, the multi-fold increase expected from the massive number of antennae will not be realized in practice. Further, in the 5G-NR downlink, each payload transmitted in the shared channel has an associated payload in the control channel. Thus, to have optimal resource utilization, the gNB should simultaneously consider the control and the shared channel payloads while allocating resources. Unlike the 4G-Long Term Evolution (4G-LTE), both control channel and shared channel support beamforming in 5G-NR. Hence, when the gNB uses the existing 4G-LTE algorithms for 5G-NR, they do not achieve the optimal resource allocation. Motivated by this, we propose a joint control and shared channel allocation for 5G-NR downlink that maximizes the sum-throughput while ensuring fairness in the allocation. We formulate this proposed resource allocation as an integer linear program. We also present low-complexity sub-optimal and approximation algorithms due to their practical usefulness. We then evaluate the proposed algorithms using system-level simulations and show that they significantly outperform the baseline algorithm. 5G New Radio Auditorium 7/29/21 15:10
Age-Of-Information Bandits with Heterogeneous Data Rates Harsh S Deshpande (IIT Bombay, India); Sucheta Ravikanti (Indian Institute of Technology, Bombay, India); Sharayu Moharir (Indian Institute of Technology Bombay, India) We consider a system with a sensor tracking a time-varying quantity and sending updates to a monitoring station using one of K different data-rates for each update. The probability of an attempted update is an unknown function of the data-rate of the update. The metric of interest is the Age-of-Information (AoI), defined as the time elapsed since the sensor made the measurement sent in the latest update received by the monitoring station. The algorithmic challenge is to determine which data-rate to use to minimize cumulative AoI over a finite time-horizon. We propose two policies and characterize their performance via analysis and simulations. One of the key takeaways is that taking the current AoI into account while determining which data-rate to use is key for good performance. In addition, we study the trade-off between AoI and throughput for the system considered. Age of Information Auditorium 7/28/21 14:30
On the Age of Information of a Queuing System with Heterogeneous Servers Anhad Bhati and Sibi Raj B Pillai (IIT Bombay, India); Rahul Vaze (TIFR Mumbai, India) An optimal control problem with heterogeneous servers to minimize the average age of information (AoI) is considered. Each server maintains a separate queue, and each packet arriving to the system is randomly routed to one of the servers. Assuming Poisson arrivals and exponentially distributed service times, we first derive an exact expression of the average AoI for two heterogeneous servers. Next, to solve for the optimal average AoI, a close approximation is derived, called the \emph{approximate AoI}, this is shown to be useful for multi-server systems as well. We show that for the optimal approximate AoI, server utilization (ratio of arrival rate and service rate) for each server should be same as the optimal server utilization with a single server queue. For two identical servers, it is shown that the average AoI is approximately $5/8$ times the average AoI of a single server. Age of Information Auditorium 7/28/21 14:50
Design of a Compact TM01-TE11 Mode Converter Using Periodic Iris Loading Ashish Chittora (BITS-Pilani, K. K. Birla Goa Campus, India) A compact TM01-TE11 mode converter design is presented in this paper. The design consists of a periodic iris loaded semicircular waveguide section and exhibits maximum conversion efficiency of 99% at 3.5 GHz. Simulated and measured results are presented for comparison and the results are in good agreement with each other. The proposed converter has a symmetric, light-weight structure and very compact design (0.37*λ) relative to earlier reported designs. The design has narrowband response and can handle up to 0.3 GW high power signal. The design is most suitable for moving platform and airborne high power microwave systems in defense applications. Antenna Design Hall A 7/29/21 16:00
Compact and Wideband Circularly Polarized Quadrature Rectangular Dielectric Resonator Antenna Abhijeet Gaonkar and Pragati Patel (NIT Goa, India) Compact and wideband Circularly Polarized (CP) Quadrature Rectangular Dielectric Resonator Antenna (QRDRA) is designed and proposed using coaxial feed. The square shaped slots filled with air epsilon r = 1 are introduced to obtain wide impedance bandwidth by reducing effective dielectric constant. The Edge grounding technique is used to miniaturize RDRA to the size 0.2 λc × 0.2 λc × 0.2 λc at frequency of 5.3 GHz. CP fields are obtained by optimizing feed location at an offset along x and y axis to excite T Ex 111 and T Ey 111 orthogonal modes. The Proposed CP Q-RDRA provides wider 3-dB axial ratio bandwidth (ARBW) ARBW ≤ 3dB of 42% (4.5-6.8 GHz), impedance bandwidth S11 ≤ −10dB of 72% (3.5-7.3 GHz) respectively. The proposed structure offers LHCP gain of more than 6 dB and radiation efficiency of more than 95% in the complete frequency range of operation. Proposed CP antenna is applicable for C-band, Wi-MAX, WLAN applications Antenna Design Hall A 7/29/21 16:20
A Compact Reconfigurable Slot-Loaded Printed Antenna for Future Wireless Applications Divyanshu Bhardwaj (Indian Institute of Information Technology Guwahati, India); Anamiya Bhattacharya (Indian Space Research Organization, India); Bidisha DasGupta (Indian Institute of Information Technology, India) In this article, one new frequency reconfigurable antenna is presented. The antenna geometry comprises an annular ring shaped slot loaded rectangular patch antenna and two p-i-n diode switches. The antenna operates over 2.4 -8.4 GHz (S band, C band, and partially in X band). The frequency hopping over the mentioned bandwidth is possible by changing switching p-i-n diode states. The proposed antenna can be used for future wireless applications such as an electromagnetic sensor for 5G and also for Cognitive Radio applications. Antenna Design Hall A 7/29/21 16:40
An H-Plane Multi-Horn Antenna Using Substrate Integrated Waveguide Technique Anil Nayak, Mr, Vinit Yadav and Amalendu Patnaik (IIT Roorkee, India) In this paper, a new H-plane multi-horn antenna is designed for small nonmetallic unmanned aerial vehicle (UAV) application. Four H-plane horn antennas are integrated into a single square substrate using the substrate integrated waveguide (SIW) concept. In fact, they are placed concentrically and directed at the four edges of the substrate. In order to control the resonant frequency at 5.8 GHz and to obtain proper matching, the co-axial connector is placed the center of the structure. The antenna provides the quasi-omnidirectional radiation instead of the directional radiation pattern at the H-plane. The laboratory prototype of the structure is measured to validate the theoretical results. This antenna is suitable for UAV applications. Antenna Design Hall A 7/29/21 17:00
Binaural Reproduction of HOA Signal Using Sparse Multiple Measurement Vector Projections Gyanajyoti Routray (Indian Institute of Technology Kanpur, India); Priyadarshini Dwivedi (Indian Institute of Technology, Kanpur, India); Rajesh M Hegde (Indian Institute of Technology Kanpur, India) Higher order Ambisonics (HOA) is one of the most promising technology in the reproduction of spatial audio in terms of spatial resolution. However binaural reproduction of spatial audio is ubiquitously used in several popular applications like AR. A novel method for binaural reproduction of HOA signals using sparse plane wave expansion is proposed in this paper. Unlike the parametric methods, the proposed method does not require prior information about the number of the discrete sources. The plane wave expansion of the encoded signals is obtained in the spherical harmonics domain using the multiple measurement vector projections, while upscaling of the input encoded signal is done to preserve the spatial resolution. Head-related transfer function (HRTF) cues are subsequently used to develop the binaural decoder. Unlike the virtual loudspeakers based approach it provides more accuracy in terms of spatial resolution as it removes the diffuse component. The efficacy of this method is illustrated using objective and subjective evaluations. Audio Processing Hall B 7/30/21 9:00
Robustness and Accuracy of Time Delay Estimation in a Live Room Yegnanarayana B. (International Institute of Information Technology, Hyderabad, India); HVS Narayana Murthy B (Research Center Imarat, India); J v Satyanarayana (DRDO INDIA, India); Vishala Pannala (International Institute of Information Technology (IIIT), Hyderabad, India); Nivedita Chennupati (International Institute of Information Technology Hyderabad, India) Estimation of time delay from the received broadband signals like speech, collected at two or more spatially distributed microphones, has many applications. Methods like the cross-correlation of the signals directly and generalized cross-correlation based methods (GCC and GCC-PHAT) have been used for several years to estimate the time delay. Performance of these methods degrades due to noise, multi-path reflections, and reverberation in a practical environment, like a live room. The estimated time delay is usually robust due to the averaging effect of the delay obtained over several frames in an utterance of a few seconds. The robustness is affected if the varying time delay of a moving speaker is desired. A smaller duration for averaging results in errors in the estimation of the time delay, and a longer duration for averaging results in loss of accuracy. Since the single frequency filtering (SFF) based analysis provides an estimation of the instantaneous time delay, it is possible to study the trade off between accuracy and robustness. This paper examines this trade-off in determining the number of stationary speakers from mixed signals and in tracking a speaker moving along a straight line path and along a circular path. The results are illustrated for actual data collected in a live room. Audio Processing Hall B 7/30/21 9:20
Learning Based Method for Robust DOA Estimation Using Co-Prime Circular Conformal Microphone Array Raj Prakash Gohil (Indian Institute of Technology, Kanpur, India); Gyanajyoti Routray and Rajesh M Hegde (Indian Institute of Technology Kanpur, India) Sound source localization in 1-Dimensional(1D) and2-Dimensional(2D) is one of the most familiar problems in signal processing. Various types of microphone arrays and their geometry have been explored to find an optimal solution to this problem. The problem becomes more challenging for a reverberate and noisy environment. Localization of the source both in the azimuth and elevation increases the complexity further. In this paper, a convolutional neural network(CNN) based learning approach has been proposed to estimate the primary source in 2D space. Further, a noble co-prime circular conformal microphone (C3M)geometry has been developed for sound acquisition. The generalized cross-correlation with phase transform (GCC-PHAT)features have been extracted from the C3M recordings, which are the input features for training purposes. The experimental results show that the learning-based estimation is more robust compared to the conventional signal processing approach. The learning-based approach also explores the GCC-PHAT features and can be adapted in an adverse acoustic environment. The performance of the proposed algorithm shows significant improvement in the root mean squared error(RMSE) and mean absolute error(MAE)scores compared to the available state-of-art methods. Audio Processing Hall B 7/30/21 9:40
Frequency-Anchored Deep Networks for Polyphonic Melody Extraction Aman Kumar Sharma (IIT Kanpur & Cisco Systems, India); Kavya Ranjan Saxena (IIT, Kanpur, India); Vipul Arora (Indian Institute of Technology, Kanpur, India) Extraction of the predominant melodic line from polyphonic audio containing more than one source playing simultaneously is a challenging task in the field of music information retrieval. The proposed method aims at providing finer F0s, and not coarse notes while using deep classifiers. Frequency-anchored input features extracted from constant Q-transform allow the signatures of melody to be independent of F0. The proposed scheme also takes care of the data imbalance problem across classes, as it uses only two or three output classes as opposed to a large number of notes. Experimental evaluation shows the proposed method outperforms a state-of-the-art deep learning-based melody estimation method. Audio Processing Hall B 7/30/21 10:00
Transfer Learning-Based Automatic Detection of Acute Lymphocytic Leukemia Pradeep Kumar Das (National Institute of Technology Rourkela, India); Sukadev Meher (National Institute of Technology, Rourkela, India) In healthcare, microscopic analysis of blood-cells is considered significant in diagnosing acute lymphocytic leukemia (ALL). Manual microscopic analysis is an error-prone and time-taking process. Hence, there is a need for automatic leukemia diagnosis. Transfer learning is becoming an emerging medical image processing technique because of its superior performance in small databases, unlike traditional deep learning techniques. In this paper, we have suggested a new transfer-learning-based automatic ALL detection method. A light-weight, highly computationally efficient SqueezNet is applied to classify malignant and benign with promising classification performance. Channel shuffling and pointwise-group convolution boost its performance and make it faster. The proposed method is validated on the standard ALLIDB1 and ALLIDB2 databases. The experimental results show that in most cases, the proposed ALL detection model outperforms Xception, NasNetMobile, VGG19, and ResNet50 with promising quantitative performance. Biomedical Image Processing Hall B 7/29/21 16:00
Optimized Bio-Inspired Spiking Neural Models Based Anatomical and Functional Neurological Image Fusion in NSST Domain Manisha Das and Deep Gupta (Visvesvaraya National Institute of Technology, Nagpur, India); Petia Radeva (Universitat de Barcelona & Computer Vision Center, Spain); Ashwini Bakde (All India Institute of Medical Sciences, Nagpur, India) Fusion of complimentary anatomical and functional information present in multi-modal medical images provides improved visualization of various bodily structures and assists radiologist to infer more factual diagnostic interpretations. Inspired by the neuronal assemblies of mammal's visual cortex, spiking neural models such as dual- channel pulse coupled neural network (DCPCNN) and coupled neural P (CNP) system efficiently extract and integrate complimentary information present in the source images. But, these models have various free parameters which are set using hit and trial approach in most of the conventional fusion methods. This paper presents an optimized multi-modal medical image fusion method in non-subsampled sheartlet transform (NSST) domain wherein the free parameters of both DPCNN and CNP system are optimized using multiobjective grey wolf optimization (MOGWO). Extensive experiments are performed on various anatomical-functional images. Subjective and objective result analysis indicate that the proposed method effectively fuse important diagnostic information of the source images and also outperforms other state of the art fusion methods. Biomedical Image Processing Hall B 7/29/21 16:20
Biomedical Image Retrieval Using Multi-Scale Local Bit-Plane Arbitrary Shaped Patterns Deepamoni Mahanta, Deepika Hazarika and Vijay Kumar Nath (Tezpur University, India) A biomedical image retrieval technique using novel multi-scale pattern based feature is proposed. The introduced technique, in each scale, employs arbitrary shaped sampling structures in addition to a classical circular sampling structure in local bit-planes for effective texture description, and named as the multi-scale local bit-plane arbitrary-shaped pattern (MS-LBASP). The proposed feature descriptor first downsamples the input image into three different scales. Then the bit planes of each downsampled image are extracted and the corresponding bit-planes are locally encoded, characterizing the local spatial arbitrary and circular shaped structures of texture. The quantization and mean based fusion is utilized to reduce the features. Finally, the relationship between the center-pixel and the fused local bit-plane transformed values are encoded using both sign and magnitude information for better feature description. The experiments were conducted to test the performance of MS-LBASP. Two benchmark computer tomography (CT) image datasets and one magnetic resonance imaging (MRI) image dataset were used in the experiments. Results demonstrate that the MS-LBASP outperforms the existing relevant state of the art image descriptors. Biomedical Image Processing Hall B 7/29/21 16:40
Detection of Myocardial Infarction from 12 Lead ECG Images Ravi Kumar Sanjay Sane (Indian Institute of Technology Guwahati, India); Pharvesh Salman Choudhary and L N Sharma (IIT Guwahati, India); Samarendra Dandapat (Indian Institute of Technology Guwahati, India) Electrocardiogram(ECG) is one of the most frequently used modality by cardiologists across the globe to detect any heart function abnormalities. In hospitals, ECG results are printed on paper by the ECG machines, which then is analysed by an expert. This work proposes a one-dimensional convolutional neural network(CNN) framework for automated myocardial infarction (MI) detection from multi-lead ECG signals extracted from ECG images. The model is developed using PTB diagnostic database consisting of 148 ECGs of (MI) cases. The results verify the efficacy of the proposed method with accuracy, sensitivity and precision of 86.21%, 89.19%, and 91.30%, respectively. The work is also compared with other state-of-the-art approaches for MI detection using ECG images. Biomedical Image Processing Hall B 7/29/21 17:00
Brain Source Localization with Covariance Fitting Approaches Anchal Yadav (Indian Institute of Technology Delhi India, India); Prabhu Babu (CARE, Indian Institute of Technology, Delhi, India); Monika Aggarwal (IIT Delhi, India); Shiv Dutt Joshi (Indian Institute of Technology, Delhi, India) The techniques like fMRI, CT scans, etc are used to localize the activity in the brain. Though these techniques have a high spatial resolution they are very expensive and uncomfortable for the patients. On the other hand, EEG signals can be obtained quite comfortably but suffer from low spatial resolution. A lot of research is being done to effectively extract spatial information from EEG signals. Many inverse techniques like MNE, LORETA, sLORETA, etc are available. All these methods can detect only a few sources and their performance degrades at low SNR. In this paper, covariance-based methods are used to estimate the location of brain activity from EEG signals such as SPICE (sparse iterative covariance-based estimation), and LIKES (likelihood-based estimation of sparse parameters). Intense simulation work has been presented to show that the proposed methods outperform the state-of-the-art methods. Biomedical Signal Processing & Networks Hall B 7/30/21 14:00
Performance Analysis of Convolutional Neural Network Based EEG Epileptic Seizure Classification in Presence of Ocular Artifacts Payal Patel (IIT Bhubaneswar, India); Udit Satija (Indian Institute of Technology (IIT), India) Recently, convolutional neural network (CNN) has played a crucial role in classifying epileptic seizures due to its capability of automatically learning the discriminatory features from the raw electroencephalogram (EEG) data. Moreover, most of the existing methods considered artifact-free EEG data for extracting features. In this paper, we analyze the impact of ocular artifacts on the performance of CNN in extracting reliable features from the EEG data for seizure classification. Furthermore, we also analyze the robustness of CNN in determining the accurate and reliable features not only from raw EEG data but also from spectral domain EEG data. The performance of the method is evaluated on the EEG signals taken from the Bonn's dataset with different types and levels of ocular artifacts. Performance evaluation results demonstrate that the classification accuracy of the method is degraded significantly under the presence of ocular artifacts. Furthermore, it is observed that the proposed CNN architecture is able to extract the discriminatory features from spectral EEG data more accurately as compared to the raw temporal EEG data. Biomedical Signal Processing & Networks Hall B 7/30/21 14:20
Implementation of a Spiking Neuron in CMOS Iman Burman (Indian Institute of Technology Kharagpur, India); Archita Hore (IIT Kharagpur, India); Ayan Chakraborty (Indian Institute of Technology Khragpur, India); Sharba Bandyopadhyay (IIT Kharagpur, India); Saswat Chakrabarti (G. S. Sanyal School of Telecommunications & Indian Institute of Technology, Kharagpur, India) A spiking neuronal network consumes very low power for computation contrary to conventional VonNeumann architectures. A CMOS based circuit which includes several features of a spiking neuron closely, is presented in this paper. Features such as refractory period, spike height and width, resting potential, spiking threshold, spike frequency adaptation and inter spike interval (ISI) have been incorporated in the circuit. A small set of parameters, chosen carefully control these features in the circuit response. The spiking pattern of the proposed circuit has been matched with selected experimental data of real biological neurons from Allen Institute for Brain Science (AIBS) database. Biomedical Signal Processing & Networks Hall B 7/30/21 14:40
Heart Rate Estimation from RGB Facial Videos Using Robust Face Demarcation and VMD Arya Deo Mehta and Hemant Sharma (National Institute of Technology, Rourkela, India) The recent studies suggest the feasibility of accessing crucial health parameters through contactless means with an RGB camera placed at a distance. As high-quality RGB cameras are getting more cost-effective due to the drastic evolution in imaging technology, the camera-based health monitoring is evoking a considerable interest among researchers. This development may provide a better alternative to the conventional contact-based methods, as it promises a convenient and contactless long term vital sign monitoring solution that doesn't restrict personal mobility. This paper introduces an effective approach towards monitoring heart rate (HR) from facial videos using an RGB camera in wild practical scenarios. The proposed approach introduces the face symmetry-based quality scoring, which is an essential step to ensure quality face detection and avoid false face detections in videos captured in a practical scenario. Further, steps such as feature points generation for optimum masking and variational mode decomposition (VMD) based filtering assist in obtaining a signal dominated mainly by the HR component. Two publicly available datasets comprising the video signals at different frame rates collected from the subjects with diverse ethnicities and skin tones are used to access the performance of the technique. The proposed approach achieved a mean absolute error of 6.58 beats per minute (BPM) on the COHFACE (Good illumination) dataset class, 9.11 BPM on the COHFACE (Bad illumination) dataset class and 6.37 BPM on the DEAP dataset class outperforming some of the state-of-art methods affirming its effectiveness in the estimation of HR in more realistic scenarios. Biomedical Signal Processing & Networks Hall B 7/30/21 15:00
Numerically Computable Lower Bounds on the Capacity of the ..(1,\infty)..-RLL Input-Constrained Binary Erasure Channel V. Arvind Rameshwar (Indian Institute of Science, Bengaluru, India); Navin Kashyap (Indian Institute of Science, India) The paper considers the binary erasure channel (BEC) with the inputs to the channel obeying the ..(1,\infty)..-runlength limited (RLL) constraint, which forbids input sequences with consecutive ones. We derive a lower bound on the capacity of the channel, by considering the mutual information rate between the inputs and the outputs when the input distribution is first-order Markov. Further, we present a numerical algorithm for numerically computing the lower bound derived. The algorithm is based on ideas from stochastic approximation theory, and falls under the category of two-timescale stochastic approximation algorithms. We provide numerical evaluations of the lower bound, and characterize the input distribution that achieves the bound. We observe that our numerical results align with those obtained using the sampling-based scheme of Arnold et al. (2006). Furthermore, we note that our lower bound expression recovers the series expansion type lower bound discussed in Corollary 5 of Li and Han (2018). We also derive an alternative single-parameter optimization problem as a lower bound on the capacity, and demonstrate that this new bound is better than the linear lower bound shown in Li and Han (2018) and Rameshwar and Kashyap (2020), for ..\epsilon > 0.77.., where ..\epsilon.. is the erasure probability of the channel. Channel Capacity and Fixed Point Analysis Hall A 7/30/21 9:00
Commitment over Compound Binary Symmetric Channels Anuj Kumar Yadav (Indian Institute of Technology Patna, India); Manideep Mamindlapally and Amitalok J. Budkuley (Indian Institute of Technology Kharagpur, India); Manoj Mishra (National Institute of Science Education and Research, Bhubaneswar, Homi Bhabha National Institute, India) In the commitment problem, two mutually distrustful parties Alice and Bob interact in a two-phase protocol, viz., commit and reveal phase, to achieve commitment over a bit string that Alice possesses. The protocol successfully achieves commitment if, firstly, Alice can commit to sharing a string with Bob, with the guarantee that this string remains hidden from Bob until she chooses to reveal it to him. Secondly, when Alice does reveal a string, Bob is able to detect precisely whether the revealed string is different from the one Alice committed to sharing. Information-theoretically secure commitment is impossible if Alice and Bob communicate only noiselessly; however, communication using a noisy channel can be a resource to realize commitment. Even though a noisy channel may be available, it is possible that the corresponding channel law is imprecisely known or poorly characterized. We define and study a compound-binary symmetric channel (compound-BSC) which models such a scenario. A compound-BSC is a BSC whose transition probability is fixed but unknown to either party; the set of potential values which this transition probability can take, though, is known to both parties a priori. In this work, we completely characterize the maximum commitment throughput or commitment capacity of a compound-BSC. We provide an optimal, computationally-efficient scheme for our achievability, and we derive a converse for general alphabet compound DMCs, which is then specialized for compound-BSCs. Channel Capacity and Fixed Point Analysis Hall A 7/30/21 9:20
Capacity of Photonic Erasure Channels with Detector Dead Times Jaswanthi Mandalapu (Indian Institute of Technology, Madras, India); Krishna P Jagannathan (Indian Institute of Technology Madras, India) We consider a photonic communication system wherein the photon detector suffers a random 'dead time' following each successful photon detection. If subsequent photon arrivals occur during the dead time, the information contained in the photons is assumed to be erased. We refer to such channels as photonic erasure channels and derive fundamental limits on the rate at which classical information can be transmitted on such channels. We assume photon arrivals according to a Poisson process, and consider two classes of detectors - paralyzable and nonparalyzable. We derive explicit expressions for the capacity of photonic erasure channels, for any general distribution of the dead times of the detector. For a photonic erasure channel with a nonparalyzable detector, we show that the capacity depends only on the expected dead time. On the other hand, with a paralyzable detector, the channel capacity depends on the dead time distribution through its Laplace transform. Channel Capacity and Fixed Point Analysis Hall A 7/30/21 9:40
The Four Levels of Fixed-Points in Mean-Field Models Sarath Yasodharan and Rajesh Sundaresan (Indian Institute of Science, India) The fixed-point analysis refers to the study of fixed-points that arise in the context of complex systems with many interacting entities. In this expository paper, we describe four levels of fixed-points in mean-field interacting particle systems. These four levels are (i) the macroscopic observables of the system, (ii) the probability distribution over states of a particle at equilibrium, (iii) the time evolution of the probability distribution over states of a particle, and (iv) the probability distribution over trajectories. We then discuss relationships among the fixed-points at these four levels. Finally, we describe some issues that arise in the fixed-point analysis when the system possesses multiple fixed-points at the level of distribution over states, and how one goes beyond the fixed-point analysis to tackle such issues. Channel Capacity and Fixed Point Analysis Hall A 7/30/21 10:00
A Novel Clustering Tendency Assessment Algorithm for WSN Generated Spatio-Temporal Data Kartik Vishal Deshpande (Indian Institute Of Technology, India); Dheeraj Kumar (IIT Roorkee, India) An algorithm is developed to visually estimate the number of clusters in wireless sensor network (WSN) generated Spatio-temporal (ST) data by separating clusters with non-contiguous sets into multiple contiguous groups in space and time (using non-contiguous to contiguous visual assessment of clustering tendency algorithm (nccVAT)). The proposed algorithm is compared with ST-DBSCAN, ST-OPTICS, and a base algorithm that applies clusiVAT to all the features. To validate nccVAT, we compare Dunn's indices of the clustering generated by various algorithms on a real-life Intel Berkeley Research Laboratory (IBRL) data-set. The algorithm with the highest Dunn's index value is accepted to provide better clusters. Clustering Hall B 7/30/21 12:00
Dimension Reduction and Clustering of Single Cell Calcium Spiking: Comparison of t-SNE and UMAP Suman Gare, Soumita Chel and Manohar Kuruba (IIT Hyderabad, India); Soumya Jana (Indian Institute of Technology, Hyderabad, India); Lopamudra Giri (IIT Hyderabad, India) Time-lapse fluorescent imaging of cytosolic calcium is used to detect cellular activity during preclinical experiments and drug screening studies. However, visualization and analysis of high dimension time series data remain challenging due to the presence of underlying heterogeneity. In this context, we propose t-distribution stochastic neighborhood embedding (t-SNE) and uniform manifold projection and approximation (UMAP) for visualization and analysis. Next, we show the density-based spatial clustering of applications with noise (DBSCAN) can be used to detect various spiking patterns present in calcium dose-response. The proposed framework combining t-SNE and DBSCAN was used to find repeating patterns, detect outliers, and label similar instances present in biological signaling. Clustering Hall B 7/30/21 12:20
An All-Digital Wideband OFDM-Based Frequency-Hopping System Using RF Sampling Data Converters Amit Sravan Bora, Harishore Singh Tourangbam and Po-Tsang Huang (National Chiao Tung University, Taiwan) Traditionally, a wideband frequency hopping spread-spectrum (FHSS) system involves the use of analog mixers and oscillators for hopping the baseband signal to different higher frequencies. Such typical heterodyne receiver architectures have higher cost and form factors. Motivated by the recent development of direct radio-frequency (RF) data converters, this paper proposes an all-digital wideband orthogonal frequency division multiplexing (OFDM) based FHSS system using an RF direct sampling receiver architecture. The frequency-hopping is done in a two-stage process, with one at the baseband and the other at the carrier frequency to make it robust from any malicious attacks in the form of eavesdropping and jamming. The all-digital process of RF signal generation and detection in the system is verified over-the-air by using the RF sampling data converters of the Xilinx Ultrascale ZCU111 RFSoC board and Qorvo RF front-end. Later, a simulated bit error rate (BER) analysis of the system under a slow-fading channel and pilot-based channel estimation is carried out, which shows comparable performance with that of an analog FHSS system at radio frequencies. Communication Systems Auditorium 7/28/21 12:10
Digital Predistortion Resource Optimization for Frequency Hopping Transceiver System Jaya Mishra (Indian Institute of Technology, Roorkee, India); Girish Chandra Tripathi (Indian Institute of Technology Roorkee, India); Meenakshi Rawat (IIT Roorkee, India) Frequency hopping (FH) is one of the best spread spectrum techniques for interference avoidance. Nonlinearity of PA is still a hindrance in using high efficiency modulation like QAM with FH. As dwell time is short, applying digital predistortion (DPD) to mitigate nonlinearity becomes critical. Memory Polynomial Model (MPM) based indirect learning architecture offers feasible solutions with reasonable resource utilization for FPGA implementation. Hard coded DPD in FPGA is the best possibility for FH system. It takes less time in the implementation and application of DPD. If a single DPD for the whole frequency band 105MHz (2.395GHz to 2.5GHz) is used, it will consume less FPGA resource but will not provide good result. Hard coded DPD at each hopping frequency is not possible because of limited resource of FPGA. So, a solution has been worked out to use six DPD, each DPD for 3 to 4 hopping frequency. Thus, this paper provides a real-time solution of DPD implementation for the FH system in the above band. NMSE has been used to judge the efficacy of DPD. The resource utilized and time taken has been studied in this paper. Communication Systems Auditorium 7/28/21 12:30
Capacity Analysis of Adaptive Combining for Hybrid FSO/RF Satellite Communication System Narendra Vishwakarma and Swaminathan Ramabadran (Indian Institute of Technology Indore, India) Free space optics (FSO) technology has fulfilled the needs of the gigabit capacity due to its exemplary features. Nevertheless, the FSO link is vulnerable to atmospheric turbulence, pointing errors, weather conditions like fog, snow etc. Subsequently, the more reliable radio frequency (RF) link can be used in combination with the FSO link to counteract the limitations. Therefore, a hybrid FSO/RF system is a promising solution for next-generation satellite communication (SATCOM) systems. In this context, we investigate an adaptive-combining-based switching scheme for a hybrid FSO/RF system considering both uplink and downlink SATCOM scenarios. Adaptive combining involves switching of the FSO link to maximal ratio combining (MRC) of FSO and RF links provided the operating FSO link quality becomes unacceptable for transmission. Further, in this paper, the performance of the adaptive-combining-based hybrid FSO/RF system is examined through exact and asymptotic ergodic capacity analysis. We have also determined a range of optimum switching threshold for different average signal-to-noise ratio (SNR) values of FSO and RF links to obtain optimal capacity performance. Moreover, we compare the performances of the adaptive-combining-based hybrid system with a single-link FSO system, MRC-based, and hard-switching-based hybrid FSO/RF systems. Communication Systems Auditorium 7/28/21 12:50
Development of Improved SOQPSK Based Data Transmission over Aeronautical Telemetry Link Ravindra Mohan Nigam (Aeronautical Development Agency, India); Pyari Mohan Pradhan (IIT Roorkee, India) In aeronautical telemetry Alamouti encoded Shaped Offset Quadrature Phase Shift Keying - Telemetry Group (SOQPSK-TG) modulated signal is used to resolve "Two antenna problem" due to simultaneous transmission from the two on-board antennae. Detection of this signal at the receiver requires estimation of channel impairments (channel gains, time delays and frequency offset). The Maximum Likelihood Sequence Estimation (MLSE) based decoder of Space Time Coding (STC) encoded SOQPSK-TG signal requires 512 states, which is too complex for implementation. In this paper, pulse shaping is performed on SOQPSK-TG frequency pulse to reduce the pulse duration. Pulse of length 2 bit interval is found to be approximately matching the SOQPSK-TG characteristic while reducing the decoder complexity to 8 number of states. Subsequently parameter estimation is carried out for STC encoded SOQPSK-2T by Maximum Likelihood (ML) estimation method. The performances of proposed pulse shaping functions are compared with those of SOQPSK-TG and Feher's Quadrature Phase Shift Keying (FQPSK-JR), and are found to be superior for aeronautical telemetry display and level flight operations. Communication Systems Auditorium 7/28/21 13:10
Channel Estimation and Data Detection of OTFS System in the Presence of Receiver IQ Imbalance Sapta Girish Babu Neelam (Bharat Electronics Limited & IIT Bhubaneswar, India); Pravas Ranjan Sahu (Indian Institute of Technology Bhubaneswar, India) Orthogonal time frequency space modulation (OTFS), which is very robust to doubly-dispersive channels under high mobility is an emerging waveform for 5G cellular applications. In this paper, we first derive the input-output vectorized relation of OTFS system in the delay-Doppler domain in presence of IQ imbalance. Next, we study the effects of receive IQ imbalance on the performance of OTFS system. We also study the channel estimation and data detection of OTFS system in the presence of receiver IQ imbalance using pilot based transmission. We use a two level threshold based technique for 1. Pilot aided channel estimation and for 2. IQ imbalance parameter estimation. We compare the performance analysis of receiver IQ imbalanced OTFS with the estimated one and observe that the error flooring effect is removed. Communication Theory 1 Auditorium 7/28/21 16:00
Parametric Estimation of SINR Distribution Using Quantized SINR Samples for Maximizing Average Spectral Efficiency Karthik Mohan K (IIT Kharagpur, India); Suvra Sekhar Das (Indian Institute of Technology Kharagpur, India) Spectrally efficient wireless communication systems are designed to dynamically adapt transmission rate and power by comparing the instantaneous signal to interference plus noise ratio (SINR) samples against SINR switching thresholds, which can be designed a priori using perfect knowledge of SINR distribution. Nevertheless, a priori perfect knowledge of SINR distribution is hardly feasible in any practical operating system for the following reasons. The operating condition is not stationary owing to mobility, while it is impossible to have prior knowledge of all possible operating conditions. Even if the set of operating conditions is defined, identifying the current operating scenario is not a trivial task either. Considering the above challenges, dynamic estimation of SINR distribution is one possible way out. The challenge encountered in such estimation is that only quantized values of SINR are available. Leveraging the well-accepted log-normal approximation of the signal to interference plus noise ratio (SINR) distribution, we develop a mechanism to obtain parametric estimates of the distribution of SINR using quantized data in this work. The proposed method can be used at the transmitter and the receiver in the same manner with appropriate modifications to signalling protocols and algorithm parameter values. We demonstrate through numerical analysis that the proposed method can help achieve near-ideal average spectral efficiency (ASE). Communication Theory 1 Auditorium 7/28/21 16:20
Optimal Pilot Design for Data Dependent Superimposed Training Based Channel Estimation in Single/Multi Carrier Block Transmission Systems Manjeer Majumder (IIT KANPUR, India); Aditya K Jagannatham (Indian Institute of Technology Kanpur, India) This paper develops a novel data dependent superimposed training technique for channel estimation in generic block transmission (BT) systems comprising of single/multicarrier (SC/MC) and zero-padded (ZP)/ cyclic prefix (CP) systems. The training sequence comprises of the summation of a known training sequence and a data-dependent sequence that is not known to the receiver. A unique aspect of the scheme is that the channel estimation is not affected by the use of a data dependent sequence. The pilot design framework is conceived in order to minimize the Bayesian Cram´er-Rao bound (BCRB) associated with channel estimation error. Simulation results are provided to exhibit the performance of the proposed scheme for single and multi carrier zero-padded and cyclic prefixed systems. Communication Theory 1 Auditorium 7/28/21 16:40
Superimposed Pilot Based Channel Estimation for MIMO Coded FBMC Systems Murali Krishna Pavuluri and Seeram Ram Prakash Sri Sai (Indian Institute of Technology Bombay, India); Vikram M. Gadre (IIT Bombay, India); Aditya K Jagannatham (Indian Institute of Technology Kanpur, India) In this paper superimposed pilot based channel estimation technique is proposed for MIMO coded FBMC systems. In coded FBMC intrinsic interference is mitigated by using spreading of symbols across time. Superimposed pilot based channel estimation is a technique that improves the spectral efficiency by transmitting the pilot symbols along with the data symbols on a set of selected subcarriers. The proposed system achieves the same bit-error performance and mean square error performance as that of MIMO-OFDM with an additional advantage of improved spectral efficiency. The spectral efficiency is improved in two ways. By removing the cyclic prefix overhead and by avoiding the use of dedicated subcarriers for pilots. Communication Theory 1 Auditorium 7/28/21 17:00
Angle of Arrival Distribution for Coherent Scattering from an Undulating Sea Surface Manishika Rawat, Brejesh Lall and Seshan Srirangarajan (Indian Institute of Technology Delhi, India) In this work, we aim to evaluate the statistical characterization of the angle of arrival (AoA) at the receiver due to coherent scattering from a random sea surface. We represent the sea surface as a Sum of Sinusoids (SoS) and model it using Pierson-Moscovitz (PM) sea wave spectrum. We evaluate the random behavior of potential scatterers along the sea surface, sea surface wave height, and their possible impact on the distribution of AoA at the receiver. Initially, analysis is carried out for a single realization of the sea surface, i.e., a sinusoidal surface. The results obtained for a sinusoidal surface are averaged to evaluate the characteristics of the ensemble-averaged SoS surface. The AoA model proposed in this work can be applied to diverse environmental conditions. The PDF so obtained can further be used to evaluate the Doppler spread and Autocorrelation function in an UW channel. Communication Theory 2 Auditorium 7/28/21 17:30
Robust Linear Transceiver Design for Parameter Tracking in IoT Networks Mohammad Faisal Ahmed (Cisco Systems (India)); Kunwar Pritiraj Rajput and Aditya K Jagannatham (Indian Institute of Technology Kanpur, India) This work develops a robust linear joint transceiver design framework toward tracking a time varying parameter in a multi-sensor network considering channel state information (CSI) uncertainty. To begin with, an optimal parameter tracking framework is developed for a scenario with perfect CSI. This is followed by formulation of the per slot average mean square error (MSE) optimization problem subject to individual sensor power constraints considering stochastic CSI uncertainty. Next, a fast block coordinate descent (BCD) based robust transceiver design is developed that minimizes the average MSE in each slot. Simulation results demonstrate the performance of the proposed scheme and also show the improvement against the existing schemes in the literature that ignore the CSI uncertainty. Communication Theory 2 Auditorium 7/28/21 17:50
Performance of Two Stage Cooperative NOMA Transmission for Full-Efficient Three User Network Ankur Bansal (Indian Institute of Technology Jammu, India); Sudhakar Modem (Indian Institute Of Technology, Jammu, India) In this paper, we investigate the performance of three user network with two-stage non-orthogonal multiple access (NOMA) scheme. In particular, the outage performance is analyzed by considering full efficiency information transmission protocol, where each user receives its own data symbol in every slot. We investigate outage performance metric at each user, and obtained closed-form expressions. Moreover, a novel fragmental outage probability at near users is evaluated, which is useful for obtaining optimum NOMA power allocation coefficients. The result shows that the considered scheme outperform conventional NOMA and optimal NOMA parameter selection maximizes the system performance. Simulation results corroborate the derived analytical expressions. Communication Theory 2 Auditorium 7/28/21 18:10
Safe Sequential Optimization in Switching Environments Durgesh Kalwar (Indian Institute of Space Science and Technology, India); Vineeth Bala Sukumaran (Indian Institute of Space Science and Technology, Trivandrum, India) We consider the problem of designing a sequential decision making agent to maximize an unknown time-varying function which switches with time. At each step, the agent receives an observation of the function's value at a point decided by the agent. The observation could be corrupted by noise. The agent is also constrained to take safe decisions with high probability, i.e., the chosen points should have a function value greater than a threshold. For this switching environment, we propose a policy called Adaptive-SafeOpt and evaluate its performance via simulations. The policy incorporates Bayesian optimization and change point detection for the safe sequential optimization problem. We observe that a major challenge in adapting to the switching change is to identify safe decisions when the change point is detected and prevent attraction to local optima. Detection & Estimation Hall B 7/30/21 12:40
Empirical Study of Weight Initializations for COVID-19 Predictions in India Meenal Narkhede, Shubham Mane, Prashant Bartakke and M Sutaone (College of Engineering, Pune) The first case of the novel Coronavirus disease (COVID-19) in India was recorded on 30th January 2020 in Kerela and it has spread across all states in India. The prediction of the number of COVID-19 cases is important for government officials to plan various control strategies. This paper presents a weekly prediction of cumulative number of COVID-19 cases in India. A graded lockdown feature, which describes the status of lockdown, is derived and incorporated in the input dataset as one of the features. For prediction, this paper proposes a model which is a stacking of different deep neural networks which have recurrent connections. Vanishing gradients is a common issue with such networks with recurrent connections. Proper weight initialization of the network is one of the solutions to overcome the vanishing gradients problem. Hence, the weight distributions and convergence performance of some state-of-the-art weight initialization techniques have been analyzed in this paper. The proposed model is initialized with the technique which would aid to avoid the vanishing gradients problem and converge faster to a lower loss. This paper also provides a comparison of the proposed model for univariate and multivariate prediction with other prediction models such as statistical model - Auto-Regressive Integrated Moving Average (ARIMA), and deep learning architectures long short term memory (LSTM), bidirectional LSTM (bi-LSTM) and gated recurrent unit (GRU). The results demonstrate that the proposed model gives better prediction results than these models. Detection & Estimation Hall B 7/30/21 13:00
Fast DFT Computation for Signals with Spectral Support Charantej Reddy Pochimireddy (IIT Hyderabad, India); Prabhu Tej V s s and Aditya Siripuram (Indian Institute of Technology Hyderabad, India) We consider the problem of computing the Discrete Fourier transform (DFT) of an N- length signal which has only k non-zero DFT coefficients at known locations. We say that a set of indices is spectral if there exists a DFT submatrix (square) with those columns that is unitary up to scaling. When the DFT support set is spectral and N is a prime power, we prove that this can be done in O(klogk) operations using k samples provided the DFT support. This is a generalization of a similar recent result for the case when N is a power of 2. DSP Algorithms Hall B 7/30/21 16:40
Improved Hankel Norm Criterion for Interfered Nonlinear Digital Filters Subjected to Hardware Constraints Srinivasulu Jogi (IIITDM Kancheepuram, India); Priyanka Kokil (Indian Institute of Information Technology Design and Manufacturing, Kancheepuram, India) This article considers the global stability analysis of interfered nonlinear digital filtering schemes implemented with fixed-point arithmetic. The proposed approach uses Hankel norm to verify the reduction of undesired memory effects of previous inputs (ringing) on future responses in nonlinear digital filters with saturation overflow nonlinearity and external disturbance. Also, the proposed criterion verifies the asymptotic stability of nonlinear digital filter without external disturbance. With the obtained results, it is shown that the suggested criterion is less restrictive than the existing criterion in the literature. By using Lyapunov stability theory, sector-based saturation nonlinearity, and Lipschitz continuity, the approach is framed in linear matrix inequality (LMI)-constraints. The efficacy, validity, and reduced conservatism of presented criterion are tested with two numerical examples. DSP Algorithms Hall B 7/30/21 17:00
Maximizing the Throughput of an Energy Harvesting Transmitter in the Presence of a Jammer with Fixed Energy Haseen Rahman (Indian Institute of Technology Bombay, India) Maximizing the data throughput of point-to-point transmitting nodes which harvest exogenous energy is a widely considered problem in literature. In this work, we consider an additive white Gaussian noise channel in the presence of a jamming adversary. The legitimate transmitter is an energy harvesting (EH) node which attempts to maximize the amount of data conveyed before a specified deadline. The jamming node, on the other hand, tries to minimize the transmitter's data throughput by introducing targeted noise. We assume that the jammer has some fixed amount of energy for interfering. When both the nodes know the EH process in advance, known as the offline setting, we compute the actions of each node at the min-max equilibrium. In the online setting, where the energy arrivals are known in a causal manner, we first consider the case without jamming and show that a simple conservative algorithm can achieve at least a quarter of the optimal offline throughput. We then show that the algorithm has the same competitiveness in the presence of an offline jammer as well. Energy Harvesting Hall A 7/30/21 10:30
Relay-Aided Bidirectional Communications Between Devices in a Hybrid User Scenario for IoT Shivam Gujral (Indian Institute of Technology Mandi, India) This paper explores a relay aided bidirectional communications scenario between two users embedded with two different technologies. One of these two energy constrained users is a backscatter device and the other is an energy harvesting (EH) device. The relay node controls the communication process between the two users in such a way that it facilitates in both energy and information cooperation and therefore, acts as a global controller for the model under consideration. Under this setting, we aim to maximize the weighted sum-throughput over a joint set of constraints in the time allocation parameter and the energy and information beamforming vectors. Henceforth, we present an optimal solution for the special case of our problem such that the relay node is equipped with a single antenna. In addition to this, we also present a sub-optimal solution to the generalized case for the multi-antenna relay node. Finally, the numerical simulations demonstrate the evaluation of our system's performance when we vary one of the key parameters of our simulation setting. Energy Harvesting Hall A 7/30/21 10:50
On Performance of Battery-Assisted SWIPT with Incremental Relaying and Nonlinear Energy Harvesting Kamal Agrawal (Indian Institute of Technology Delhi, India); Anand Jee and Shankar Prakriya (Indian Institute of Technology, Delhi, India) This paper investigates the performance of incremental relaying (IR) in a two-hop network with a battery-assisted EH relay. Assuming nonlinear energy harvesting (EH) and time-switching protocol at relay, expressions are derived for outage probability and throughput in closed-form to show that augmenting the harvested energy by a small amount of battery energy significantly enhances the throughput. We demonstrate that unlike linear EH, nonlinear EH causes loss of diversity, making use of the direct link very important. Further, using the asymptotic expression for the outage probability, we establish concavity of throughput with respect to EH parameter and demonstrate that a judicious choice of the TS parameter is essential in order to maximize the throughput when fixed battery energy is available per signalling interval. Judicious choice of target rate is important to minimize battery energy consumption at the relay. Monte Carlo simulations confirm accuracy of the derived expressions. Energy Harvesting Hall A 7/30/21 11:10
Automated Macular Disease Detection Using Retinal Optical Coherence Tomography Images by Fusion of Deep Learning Networks Latha V and Ashok R (College of Engineering, Trivandrum, India); Sreeni K G (College of Engineering, Thiruvananthapuram, INDIA, India) This work proposes a method to improve the automated classification and detection of macular diseases using retinal Optical Coherence Tomography (OCT) images by utilizing the fusion of two pre trained deep learning networks. The concatenation of feature vectors extracted from each of the pre trained deep learning model is performed to obtain a long feature vector of the fused network. The experimental results proved that the fusion of two Deep Convolution Neural Network (DCNN) achieves better classification accuracy compared to the individual DCNN models on the same dataset. The automated retinal OCT image classification can assist the large-scale screening and the diagnosis recommendation for an ophthalmologist. Image Processing Hall B 7/29/21 14:30
Selective Variance Based Kinship Verification in Parent's Childhood and Their Children Madhu Oruganti (National Institute of Technology RAIPUR, India); Toshanlal Meenpal (NIT Raipur, India); Saikat Majumder (National Institute of Technology Raipur, India) Based on two facial image appearances estimating their kinship is the main aim of the kinship verification. Since a decade of time this problem is attempted by many sophisticated algorithms but still many challenging dimensions are unsolved. Among these, age progression-based kinship verification is one of the obscure parts. The similarities in facial features between parent and their children will be numerous in their childhood. As age progress, child facial features are varied and dispersed from parent facial features. It becomes a challenging task to estimate their kinship. So, a new dimensional database with parent in childhood and their child images is collected. This paper proposes and trains a metric to ensure that the model can predict whether the given pair images are kin or non-kin. In training module, differences of Histogram of Gradient (HoG) features for all combinations of pairs are computed and each pair absolute differences are calculated. Further, selective minimum variances are used to assess the kin similarity features. A global threshold is computed to classify kins and non-kins. After this comprehensive training, testing is also done in a similar way. The computed global threshold in training module is effectively used to estimate kinship verification in testing module. Experimental results are presented and out performed with an accuracy of 82%. Image Processing Hall B 7/29/21 14:50
Attention-Based Phonetic Convolutional Recurrent Neural Networks for Language Identification Ramesh Gundluru (Indian Institute of Technology, Hyderabad, India); Vayyavuru Venkatesh (Indian Institute of Technology Hyderabad, India); Kodukula Sri Rama Murty (Indian Institute of Technology, Hyderabad, India) Language identification is the task of identifying the language of the spoken utterance. Deep neural models such as LSTM-RNN with attention mechanism shown great potential in language identification. The language cues like phonemes and their co-occurrences are an important component while distinguishing the languages. The acoustic feature-based systems do not utilize phonetic information. So the phonetic feature-based LSTM-RNN models have shown improvement over the raw-acoustic features. These methods require a large amount of transcribed speech data to train the phoneme discriminator. Obtaining transcribed speech data for low resource Indian languages is a difficult task. To alleviate this issue, we investigate the usage of pre-trained rich resource phonetic discriminators for low resource target languages to extract the phonetic features. We then trained an attention CRNN based end-to-end utterance level language identification (LID) system with these discriminative phonetic features. We used open-source LibriSpeech English data to train the phoneme discriminator with sequence discriminate objective lattice-free maximum mutual information (LF-MMI). We achieved overall 20% absolute improvements over the baseline acoustic features CRNN model. We also investigate the significance of the duration in LID. Machine Learning Hall B 7/30/21 10:30
Spatio-Temporal Prediction of Roadside PM2.5 Based on Sparse Mobile Sensing and Traffic Information Anand Kakarla (Indian Institute Of Technology Hyderabad, India); Venkata Satish Kumar Reddy Munagala (Indian Institute of Technology Hyderabad, India); Tetsuhiro Ishizaka and Atsushi Fukuda (Nihon University, Japan); Soumya Jana (Indian Institute of Technology, Hyderabad, India) While real-time management of urban mobility has become common in modern cities, it is now imperative to attempt such management subject to a sustainable emission target. To achieve this, one would require emission estimates at spatio-temporal resolutions that are significantly higher than the usual. In this paper, we consider roadside concentration of PM2.5, and make predictions at high spatio-temporal resolution based on location, time and traffic levels. Specifically, we optimized various machine learning models, including ones involving bagging and boosting, and found Extreme Gradient Boosting (XGBoost, XGB) to be superior. Moreover, the tuned and optimized XGB utilizing traffic information achieved significant gain in terms of multiple performance measures over a reference method ignoring such information, indicating the usefulness of the latter in predicting PM2.5 concentration. Machine Learning Hall B 7/30/21 10:50
Early Prediction of Human Action by Deep Reinforcement Learning Hareesh Devarakonda (IIIT Sricity, India); Snehasis Mukherjee (Shiv Nadar University, India) Early action prediction in video is a challenging task where the action of a human performer is expected to be predicted using only the initial few frames. We propose a novel technique for action prediction based on Deep Reinforcement learning, employing a Deep Q-Network (DQN) and the ResNext as the basic CNN architecture. The proposed DQN can predict the actions in videos from features extracted from the first few frames of the video, and the basic CNN model is adjusted by tuning the hyperparameters of the CNN network. The ResNext model is adjusted based on the reward provided by the DQN, and the hyperparameters are updated to predict actions. The agent's stopping criteria is higher or equal to the validation accuracy value. The DQN is rewarded based on the sequential input frames and the transition of action states (i.e., prediction of action class for an incremental 10 percent of the video). The visual features extracted from the first 10 percent of the video is forwarded to the next 10 percent of the video for each action state. The proposed method is tested on the UCF101 dataset and has outperformed the state-of-the-art in action prediction. Machine Learning Hall B 7/30/21 11:10
A Deep Reinforcement Learning Approach for Shared Caching Pruthvi Trinadh and Anoop Thomas (Indian Institute of Technology Bhubaneswar, India) A client-server network in which multiple clients are connected to a single server possessing files/data through a shared error free link is considered. Each client is associated with a cache memory and demands a file from the server. The server loads the cache memory with a portion of files during off-peak hours to reduce the delivery rate during peak hours. A decentralized placement approach which is more practical for large networks is considered for filling the cache contents. In this paper, the shared caching problem in which each cache can be accessed by multiple clients is considered. A Deep Reinforcement Learning (DRL) based framework is proposed for optimizing the delivery rate of the requested contents by the users. The system is strategically modelled as a Markov decision process, to deploy our DRL agent and enable it to learn how to make decisions. The DRL agent learns to multicast coded bits from the file library of the server in such a way that the user requests are met with minimum transmissions of these coded bits. It is shown that the proposed DRL based agent outperforms the existing decentralized algorithms for the shared caching problem in terms of normalized delivery rate. For the conventional caching problem which is a special case of the shared caching problem, simulation results show that the proposed DRL agent outperforms the existing algorithms. Machine Learning Hall B 7/30/21 11:30
Learning to Decode Trellis Coded Modulation Jayant Sharma and V. Lalitha (IIIT Hyderabad, India) Trellis coded modulation (TCM) is a technique combining modulation with coding using trellises designed with heuristic techniques that maximize the minimum Euclidean distance of a codebook. We propose a neural networks based decoder for decoding TCM. We show experiments with our decoder that suggest the use of Convolutional Neural Network (CNN) with Recurrent Neural Network (RNN) can improve decoding performance and provide justification for the same. We show the generalization capability of the decoder by training it with small block length and testing for larger block length. We also test our decoder for its performance on noise model unseen in the training. Machine Learning for Communications Auditorium 7/28/21 15:10
Auto-SCMA: Learning Codebook for Sparse Code Multiple Access Using Machine Learning Ekagra Ranjan and Ameya Vikram (Indian Institute of Technology (IIT) Guwahati, India); Alentattil Rajesh (IIT G, India); Prabin Kumar Bora (Indian Institute of Technology Guwahati, India) Sparse Code Multiple Access (SCMA) is an effective non-orthogonal multiple access technique that facilitates communication among users with limited orthogonal resources. Currently, its performance is limited by the quality of the handcrafted codebook. We propose Auto-SCMA, a machine learning based approach that learns the codebook using gradient descent while using a Message Passing Algorithm decoder. It is the first machine learning based approach to generalize successfully on the Rayleigh fading channel. It is able to learn an effective codebook without involving any human effort in the process. Our experimental results show that Auto-SCMA outperforms previous methods including machine learning based methods. Machine Learning for Communications Auditorium 7/28/21 15:30
Joint Beamwidth and Number of Concurrent Beams Estimation in Downlink mmWave Communications Nancy Varshney (IIT Delhi, India); Swades De (Indian Institute of Technology Delhi, India) This paper proposes a sectored-cell framework for mmWave communication. It consists of multiple concurrent beams generated from a partially-connected hybrid precoder at an eNodeB (eNB) to serve a dense user population in urban scenarios. Multiple beams sweep the cell in a round-robin fashion to serve the sectors with fair scheduling opportunities. Each beam serves all the users located within a sector using orthogonal frequency division multiple access. We aim to estimate an optimum beamwidth and an optimum number of beams required to maximize the average of long-run user rates with a given power budget for transmission and hardware consumption at the eNB. Simulation results demonstrate that employing higher beams increases the side-lobe interference still, the achievable average long-run user rate improves on account of longer sector sojourn time and higher frequency reuse. On the other hand, employing a very narrow beam is also not optimal. Millimeter Wave Systems Auditorium 7/29/21 8:20
A Blind Iterative Hybrid Analog/Digital Beamformer for the Single User mmWave Reception Using a Large Scale Antenna Array Yash Vasavada, Naitik N Parekh and Aarushi Dhami (Dhirubhai Ambani Institute of Information and Communication Technology, India); Chandra Prakash (Space Applications Centre & ISRO, India) This paper develops a blind single-user beamforming algorithm for partially-connected hybrid analog-digital (HAD) antenna arrays operating at millimeter wave (mmWave) frequencies. The proposed scheme is optimal and computationally efficient - it is shown to converge to the optimal eigenvector beamformer without requiring an explicit eigenvector decomposition (EVD) of the received signal's correlation matrix. The Direction of Arrival (DoA) is estimated using the Fast Fourier Transform (FFT) of the digitally-formed beam weight vector and it is used to continually update the analog beams formed by the subarrays of the large array in a tracking mode. The simulation of the proposed scheme shows the bit error probability and DoA estimation performance close to the theoretical limits. Millimeter Wave Systems Auditorium 7/29/21 8:40
Multi-Source TCP (MSTCP): A Transport Protocol for Distributed Content Delivery Lalhruaizela Chhangte (IITB-Monash Research Academy, India); Pramey Singh and D. Manjunath (IIT Bombay, India); Nikhil Karamchandani (Indian Institute of Technology Bombay, India) Fetching different parts of the same content (file) simultaneously through multiple network paths has been found to improve content delivery. There are several application layer programs that use this technique to improve the perceived performance at the end users. However, these applications use multiple sockets (multiple connections) at the transport layer, which has several disadvantages. Also, the existing transport layer protocols that allow content delivery over multiple network paths over a single transport layer connection (e.g., MPTCP) are limited to content delivery from a single source. With the availability of content across distributed content servers, there is a need for a transport layer protocol that provides the ability to deliver content from these distributed sources over a single transport layer connection. In this paper, we design and implement a multi-source transport control protocol (MSTCP) that can be used to deliver content from a distributed source to a client application over a single transport layer connection. A prototype implementation and preliminary performance measures showing the effectiveness of MSTCP are also presented. Networking Hall A 7/29/21 14:30
On Traffic Classification in Enterprise Wireless Networks Sipra Behera, Bighnaraj Panigrahi and Hemant Kumar Rath (Tata Consultancy Services, India); Jyotirmoy Karjee (Samsung, Bangalore, India) Enterprises today are quickly adopting intelligent, adaptive and flexible wireless communication technologies in order to become compliant with Industry 4.0. One of the technological challenges related to this is to provide Quality of Services (QoS)-enabled network connectivity to the applications. Diverse QoS demands from the applications intimidate the underlying wireless networks to be agile and adaptive. Since the applications are diverse in nature, there must be a mechanism to learn the application types in near real-time so that the network can be provisioned accordingly. In this paper, we propose a Machine Learning (ML) based methods to classify the application traffic. Our method is different from the existing port based and Deep Packet Inspection (DPI) based methods and uses statistical features of the network traffic related to the applications. We validate the performance of the proposed model in a lab based SDNized WiFi set-up. SDNization ensures that the proposed model can be deployed in practice. Networking Hall A 7/29/21 14:50
A Light-Weight Delay Tolerant Networking Framework for Resource-Constrained Environments Ajay Salas (Indian Institute of Space Science and Technology, India); Sarath Babu (Indian Institute of Space Science and Technology (IIST), India); Manoj Bs (Indian Institute of Space Science and Technology, India) Next generation communication infrastructures are characterized by customized network environments deployed for meeting the application or user specific needs as well as for achieving the required Quality-of-Service (QoS). The surge of mobile devices and their applications form a major bottleneck in realizing the QoS due to the resource constraints in mobile devices and the uncertain mobility pattern of users. Delay/Disruption Tolerant Networking (DTN) approaches are employed to cope with the issues in dynamic wireless environments such as intermittent connectivity, high error rate and packet loss, and network heterogeneity. However, the overhead required in terms of protocol, memory, and computational power in traditional DTN approaches may not be suitable for energy-constrained mobile devices. Therefore, we propose a Light-Weight DTN (LWDTN) framework for resource-constrained delay/disruption-prone wireless environments. We follow the traditional custody-transfer approach in designing the LWDTN framework with three types of bundles involving minimal header fields. The experimental results from a DTN testbed show the efficacy of LWDTN in delivering around 80% packets within the feasible time. Networking Hall A 7/29/21 15:10
Differential Scale Based Multi-Objective Task Scheduling and Computational Offloading in Fog Networks Mohit Kumar Saxena (Indian Institute of Technology, Patna, India); Sudhir Kumar (Indian Institute of Technology Patna, India) Cloud computing suffers from various challenging issues in Internet of Things (IoT) networks like real-time response, energy-efficient execution, and cost of computation. Fog is an emerging distributed computing paradigm which is useful for delay-sensitive tasks in IoT network. An offloading strategy decides where to offload the task and a task scheduling strategy chooses an appropriate fog node based on the requirements of the task while meeting the quality of services (QoS) criteria. Although the computational offloading and task scheduling problem has been widely studied, there is very limited research on delay-energy tradeoff. We propose a fog network that follows an M/M/c queue for computational offloading and a differential scale-based Best Worst Method (BWM) for computation of optimal weights in multi-objective task scheduling. The optimization problem minimizes the execution delay while meeting QoS criteria. The numerical experiments show the efficacy for the different QoS criteria. Networking Hall A 7/29/21 15:30
Generation of Optical Frequency Comb Using Cascaded Brillouin Scattering at Low Power Utilizing Pump Recycling Technique in a Single Mode Fiber Aritra Paul and Pradeep Kumar (Indian Institute of Technology Kanpur, India) The paper describes the process of optical frequency comb generation using cascaded stimulated Brillouin scattering in optical fibers. The cascaded stimulated Brillouin scattering process is induced by the SBS-pump recycling technique in a single mode fiber. The single mode fiber is placed inside a recirculating cavity, with a loop mirror placed at the terminal end of the fiber. The pumps are obtained from four wave mixing process in a semiconductor optical amplifier. We have achieved a total of 8 comb lines − 5 lines within 6 dB power variation. The comb lines are separated by approximately 11 GHz (∼0.085 nm). Optical Communication Hall A 7/28/21 12:10
Multi-Rate Kalman Filter for Carrier Phase Recovery in 200 Gbps PDM Coherent Optical Receivers Wrivu Sanyal (Indian Institute of Technology, Kanpur, India); Srishti Sharma (Indian Institute of Technology, Delhi, India); Pradeep Kumar (Indian Institute of Technology Kanpur, India) The Kalman filter is often used for tracking and estimation of effects such as LPN and NLPN in long haul coherent optical communication systems. However, real-time symbol-by-symbol estimation of these parameters is computationally challenging. We use a multi-rate Kalman filtering scheme that allows for different sampling and state update rates in the system. This scheme achieves high Q-factor by making use of maximum available samples while reducing computational load. Simulations are performed for 200 Gbps PDM-16-QAM system by transmitting 20000 symbols over 800 km optical channel. The filter has Q-factor of 17.25 dB with state estimates being updated after every 20 samples. The filter shows more than 1 dB improvement in Q-factor when compared to a KF where the intermediate samples are not utilised for phase estimation. Optical Communication Hall A 7/28/21 12:30
Optical Sideband Interference Using Optical IQ and Mach-Zehnder Modulators Govind Kumar, Nishant Chandra and Pradeep Kumar (Indian Institute of Technology Kanpur, India) We observe optical sideband interference using an optical IQ modulator at the transmitter and a Mach-Zehnder modulator at the receiver. We measure 82% interference visibility in back-to-back configuration and 71% interference visibility over 25 km optical fiber channel. We also show the simulation of optical sideband interference using IQ modulator and Mach- Zehnder modulator. We measure 77% visibility by simulation in back-to-back connection of transmitter and receiver. We study the effect of the linewidth of the laser on the optical spectrum of the sideband. We derive expressions for sideband power as a function of the applied phase difference between transmitter and receiver in the presence of chromatic dispersion in the fiber. Optical Communication Hall A 7/28/21 12:50
Mode Analysis of AlGaAs Based Hybrid Metal Insulator Plasmonic Waveguide with Nanoscale Confinement Santosh Kumar (NIT PATNA, India); Pintu Kumar and Rakesh Ranjan (National Institute of Technology Patna, India) An aluminum gallium arsenide (AlGaAs) based hybrid metal insulator plasmonic waveguide (HMIPW) has been investigated to analyze the optical properties of fundamental and higher order modes, such as real part of effective index, normalized effective mode area, and propagation length at 1550 nm of wavelength. The modal investigations have been done by varying the thicknesses of high index, low-index, and metal regions. The propagation length, up to 480 μm has been achieved, for the fundamental mode propagation. The multi-mode analysis presented in the current work can be extended for the analysis of biosensors, multimode interferometer, high speed optical signal processing, etc. Optical Communication Hall A 7/28/21 13:10
A Penalty-Based Routing and Spectrum Assignment in Fragmented Elastic Optical Network Spectrum Anjali Sharma and Sobir Ali (Indian Institute of Technology Kanpur, India); Varsha Lohani (IIT Kanpur, India); Yatindra Nath Singh (Indian Institute of Technology Kanpur, India) Routing and spectrum assignment (RSA) has been an area of keen interest in Elastic Optical Networks (EONs). Improper resource provisioning causes fragmentation in the network spectrum, which leads to inefficient spectrum utilization. It also causes an increase in blocking of the new connection requests. Fragmentation management techniques are complicated and costly. There is a need to operate the network in a fragmented state without worsening the performance. In this work, we present a penalty-based routing and spectrum assignment technique to mitigate fragmentation effects. We also propose a best-effort routing and spectrum assignment if the demanded spectrum resources are not available. The simulation results show that the proposed techniques perform better in terms of resource blocking ratio and network spectrum utilization. Optical Systems Hall A 7/28/21 14:30
Neural Networks for Predicting Optical Pulse Propagation Through Highly Nonlinear Fibers Naveenta Gautam (Indian Institute of Technology, India); Amol Choudhary (Indian Institute of Technology (IIT), Delhi, India); Brejesh Lall (Indian Institute of Technology Delhi, India) Due to increase in demand of the optical fiber communication system there is a special emphasis on diagnosing ultrashort pulses. The linear and nonlinear distortions introduced during transmission gives rise to wide variety of wave dynamics. The conventional signal processing techniques being used for characterising these pulses are computationally inefficient. Since machine learning has shown improvement compared to other analytical methods, we present a comparative study of different neural network (NN) architectures to predict the output pulse profile after transmission through highly nonlinear and dispersive fibers. The trained network has the ability to learn the mapping from a set of input and output pulses for the case of both known and unknown fibers. Since each NN has its own advantages and disadvantages, we to the best of our knowledge, present a comprehensive analysis of six different NN architectures (i) fully connected NN (FCNN), (ii) cascade forward NN (CaNN), (iii) Convolutional NN (CNN), (iv) long short term memory network (LSTM), (v) bidirectional LSTM (BiLSTM) and (vi) gated recurrent unit (GRU) for the first time. Optical Systems Hall A 7/28/21 14:50
Coupling Length Analysis by Using Three Strips for Compact Design of Photonic Waveguide in Photonic Integrated Circuit Veer Chandra (National Institute of Technology Patna, India); Santosh Kumar (NIT PATNA, India); Rakesh Ranjan (National Institute of Technology Patna, India) Development of compact photonic waveguide makes the possibility to design very dense photonic integrated circuits by using complex design structure. Conversely, compact photonic waveguide requires high coupling length between the neighboring photonic waveguides to avoid crosstalk between them. Some recent research demonstrated that to increase the coupling length between adjacent photonic waveguides, one can use the strips between them. In the current work, our focus is to obtain higher coupling length with reduced separation between the adjacent waveguides. By using three uniform silicon strips, the maximum coupling length of 3029 µm has been achieved for the end-to-end separation between photonic waveguides of 300 nm. The obtained result is higher than previously reported values at the same separation of photonic waveguides. Higher coupling length is beneficial to design many compact photonic devices such as splitter, photonic switches, etc. Optical Systems Hall A 7/28/21 15:10
Quantized Feedback-Based Space Shift Keying in Visible Light Communication Sivanjan Rao Chandhu (Indian Institute of Technology, Roorkee, India); Anshul Jaiswal (Indian Institute of Technology Roorkee, India) This work proposes a novel feedback-based space shift keying (SSK) scheme to remove the limitations of the conventional SSK scheme in the visible light communication (VLC) system. The proposed scheme requires finite feedback bits to transfer the knowledge of channel gain ordering at the transmitter side and therefore termed as quantized feedback-based SSK (QF-SSK) scheme. Due to the symmetric nature of the channel gain in the VLC system, there are many locations where two or more channel gains become the same, which leads to the very high bit error rate (BER) performance for the SSK scheme. Furthermore, the fixed spatial constellation of the SSK scheme in the VLC system may not follow the gray mapping for real-line constellation points at every location, which is another reason for high BER. To overcome these limitations, the proposed QF-SSK scheme provides uneven power allocation between LEDs and uses an adaptive spatial constellation based on the information of channel gain order. It is noticed from the analysis that the proposed QF-SSK scheme significantly improves BER performance over all locations in VLC system as compared to conventional SSK scheme with very small feedback bits. Optical Systems Hall A 7/28/21 15:30
Slotted Aloha for FMCW Radar Multiple Access Networks Haritha K (Indian Institute of Science, India); Vineeth Bala Sukumaran (Indian Institute of Space Science and Technology, Trivandrum, India); Chandramani Singh (Indian Institute of Science, India) We study medium access in FMCW radar networks. In particular, we propose a slotted ALOHA protocol and analyze probability of interference between radars as a function of system parameters such as total number of radars, chirp duration, number of chirps in a repetition interval, as well as medium access probability. We see that the characteristics of interference probability in FMCW radar networks are very different from those in wireless communication networks. We observe that interference probability also depends on the number of chirps in a radar packet. We further propose a notion of throughput and study its variation with various parameters. We perform extensive simulation to verify our analytical results. Radar Hall B 7/29/21 8:20
Analysis of 5G New Radio Waveform as an Illuminator of Opportunity for Passive Bistatic Radar Purushottama Lingadevaru (National Institute of Technology Karnataka, India); Pardhasaradhi Bethi (National Institute of Technology - Karnataka, India); Pathipati Srihari, Dr (National Institute of Technology Karnataka, India); G. V. k. Sharma (GITAM University, India) Passive radar detects targets using the reflections of electromagnetic signals illuminated by unintended sources of opportunity in the given surveillance region. The illuminators of opportunity (IOO) like FM, DVB, DAB, LTE, WiMax, and radio frequency signals are used for the passive radar depending on the availability, frequency of operation and, type of application. This paper proposes the upcoming 5G New Radio waveform (5G NR) as an IOO for passive bistatic radar. The 5G NR waveform is used to perform parametric analysis of passive bistatic radar. The radar parameters like range resolution, velocity resolution, range product, maximum unambiguous PRF, and Cassini's ovals are investigated. Further, the 5G NR IOO is compared against existing LTE and other IOOs. Simulation results reveals that all the radar parameters are outperforming for the 5G NR waveform, claiming that 5G NR is a potential candidate for the future IOO. Radar Hall B 7/29/21 8:40
Greedy Successive Relaying for the Multicarrier Diamond Relay Channel Antony Mampilly (IIT Madras, India); Srikrishna Bhashyam (Indian Institute of Technology Madras, India) We propose a new decode-and-forward (DF) protocol called Greedy Successive Relaying (GSR) protocol for the M-relay multicarrier diamond relay channel. We consider the diamond channel with half-duplex relays. The GSR protocol uses two states and the relays are partitioned into two sets, A and B. In the first state of the GSR protocol, the relays in A will receive messages from the source while the relays in B will transmit messages to the destination. In the second state, the role of the relays will be reversed i.e. the relays in A will be transmitting to the destination while the relays in B will be receiving from the source. The GSR protocol takes advantage of:(1) successive relaying to overcome the half-duplex limitation, and (2) greedy subcarrier allocation across relays to exploit the available diversity from all the relays. The proposed GSR protocol performs significantly better than the existing protocols. Furthermore, under a Rayleigh fading model, the gap from the cutset bound is also observed to be bounded at high power. Relays Hall A 7/30/21 12:00
Outage Analysis of SWIPT-Enabled Multi-Relay Aided Full Duplex NOMA System Under Partial Relay Selection Shubham Shinkar and Babu A v (National Institute of Technology Calicut, India) This paper considers a multi-relay aided full duplex (FD) cooperative non-orthogonal multiple access (NOMA) system integrated with simultaneous wireless information and power transfer (SWIPT) technique, which consists of a base station (BS), N FD relays and two downlink NOMA users. We propose variable transmit power allocation (VTPA), where the power allocation factor at the base station (BS) and the time switching (TS) factor at the selected relay are chosen adaptively so as to improve the performance of the downlink users in the system. Assuming partial relay selection (PRS) scheme, analytical expressions are derived for the outage probabilities (OP) of the downlink users under VTPA and constant transmit power allocation (CTPA) schemes. To gain more insights, we describe the asymptotic OP (AOP) analysis as well. We show that the proposed VTPA scheme can mitigate the outage floor present in FD-NOMA system under CTPA and significantly improve the OP performance of the downlink users. Relays Hall A 7/30/21 12:20
Analysis of Multi-Hop Device-To-Device Networks with Half-Duplex Relays Gourab Ghatak (IIIT Delhi, India) In this paper, we analyze a multi-hop device-to-device (D2D) communication network operating in a region of cellular outage, e.g., in case of a natural disaster. In particular, we assume that the D2D devices operate in a half-duplex manner and can receive signals from or transmit to a single D2D relay. For this system, we characterize a multi-hop D2D transmission protocol wherein we divide the transmission area into multiple regions based on the D2D transmission range. In contrast to the other works present in literature, we have taken into account the probability of a relay being used by another D2D source at the instant when the typical D2D source attempts to connect to it. Then, we derive the signal to interference and noise ratio (SINR) coverage probability for a typical device. Based on this, we define and characterize a performance metric called the availability-coverage product (ACP) to jointly take into account the coverage performance of the devices and the probability of them being used as relays by the other devices in outage. Our analysis highlights several system design insights in terms of the D2D communication range and the optimal number of active devices in the network in terms of the ACP. Relays Hall A 7/30/21 12:40
On Outage Probability Analysis of Uplink Cooperative VFD-NOMA over Nakagami-m Faded Channels Justin Jose (Indian Institute of Technology Indore, India); Parvez Shaik (Indian Institute of Technology Indore, India); Shubham Bisen and Vimal Bhatia (Indian Institute of Technology Indore, India) Full-duplex (FD) and non-orthogonal multiple access (NOMA) communications are promising technologies which can efficiently utilize the available scarce spectrum in comparison to the traditional half-duplex (HD) and orthogonal multiple access (OMA) techniques. However, due to high residual self-interference (RSI) in the practical FD systems, virtual FD (VFD) relaying has garned appealing research attention in both academia and industry. In this work, we present a study of an uplink VFD-NOMA system with a base station, two relays, and two cell-edge uplink users, and analyze their performance over generalized Nakagami-m fading channels. Specifically, closed-form expressions of outage probability (OP) for both the users are derived for the considered system model, with the accuracy being verified through rigorous simulations. Moreover, the performance of the users is compared with the conventional FD-NOMA and FD-OMA schemes. Relays Hall A 7/30/21 13:00
Wideband Circuit Analog Absorber Using Modified Resistive Cross-Dipoles Aditya Mandar Jabade (BITS-Pilani, India); Hrishikesh Sonalikar (BITS, Pilani, India) In this paper, a novel circuit analog absorber design is proposed using resistive cross-dipoles as the basic template. The absorber consists of two FR4 substrates separated by air. The top substrate consists of a modified cross-dipole frequency selective surface (FSS) having lumped resistors and the bottom substrate provides a ground plane. The proposed absorber has a very low profile of 0.0739 wavelengths at its lowest operating frequency and a fractional bandwidth of 100.55% from 3.13 to 9.46 GHz. The designed absorber is insensitive to incident polarization and shows stable performance at oblique angles of incidence. It is demonstrated that a slight modification in the form of a rectangular strip applied to the basic cross-dipole FSS reduces the thickness to bandwidth ratio of the absorber. RF & Microwave 1 Hall A 7/28/21 16:00
Influence of Various Soil Types and Its Properties on Filamentary Planar Coil Based Magnetic Induction Communication System Swathi Sugumar and Sakthivel Murugan Santhanam (SSN College of Engineering, India) A novel idea of using compact filamentary planar spiral coils for the magnetic induction (MI) based underground (UG) communication to achieve high received power and enhanced transmission distance is proposed. In the existing system, non-planar coils were employed as transceivers which lost their function due to their huge size and deployment difficulty. An enhanced MI UG channel model is proposed to accurately investigate the UG medium's influence on the MI system performance by considering various soil properties that were considered negligible in the earlier models. An analytical approach to calculate self-inductance and mutual inductance of circular and square coils are described from which the communication parameters such as received power, path-loss, and signal-to-ratio are derived. The simulation results reveal that the square coil achieves 9.22% higher received power than the circular coil due to its high inductive area and low resistance. Further, the influence of coil parameters, soil properties, and coil misalignment on the received power is studied for the proposed filamentary planar square spiral coil (FPSSC) and it's least and most sensitive parameters are identified. The received power of the proposed FPSSC system exhibits a significant improvement of 59.46% as compared to the traditional non-planar MI coil system. RF & Microwave 1 Hall A 7/28/21 16:20
Two-Way Array Factor Supported by Thinning Strategy for an Improved Radar Performance Rathod Rajender (NIT-Rourkela, India); Konidala Subhashini (National Institute Of Technology, India); B Pavan Kumar (APSD Communication Systems Group U R Rao Satellite Centre ISRO, India) The presence of side-lobes which are adjacent to main -lobe are serious concern in radar systems. The amplitude distributions across the array aperture improves the side-lobe performance at the cost of directivity and hence aperture efficiency. The current two-way array pattern optimization techniques reveals that if the receive pattern nulls are placed along transmit pattern side-lobe peaks and by using two weight amplitude distribution, side-lobe levels (SLL) can be suppressed up to -50dB. In this work, these techniques are further investigated using three weight amplitude excitation to achieve the best design of a radar array resulting in SLL less than -57dB. These solutions has been validated by parametric optimization and supported by two case studies. RF & Microwave 1 Hall A 7/28/21 16:40
A Miniaturized Wideband Half Mode Substrate Integrated Waveguide Bandpass Filter for C Band Applications Swathy BH and Pallepogu Prasanna Kumar (IIITDM Kancheepuram, India); Prerna Saxena (Indian Institute of Information Technology, Design and Manufacturing, Kancheepuram, India) We propose a novel compact wideband bandpass filter for C band applications. The bandpass filter is designed using half mode substrate integrated waveguide and defected ground structure in the form of periodic circular vias are incorporated into it. RT/Duroid 5880 with a permittivity of 2.2 and thickness of 0.508mm is used as substrate for designing the proposed bandpass filter. Detailed parametric analysis is presented to study the influence of various design parameters on the filter characteristics. The proposed filter exhibits a passband over 5.7-7.3GHz with |S11| < -10dB and |S21| ~ 0.6dB. In addition, the filter has a 3dB fractional bandwidth of 23.2% along with a small size of 0.015 lambdag^2. The proposed bandpass filter has higher 3dB bandwidth and a much smaller size as compared to the state-of-the-art designs. RF & Microwave 2 Hall A 7/28/21 17:30
Design of an Elementary Microstrip Power Splitter for Antenna Array Chanchala Kumari (BIT MESRA RANCHI, India); Neela Chattoraj (Birla Institute of Technology, Mesra Ranchi, India) A new simple microstrip power divider is presented for an antenna array. In this proposed structure micro strip lines are used in place of conventional waveguide, so that its overall size reduced. In this paper the two-power splitter are presented one is the simple T junction and other one is 1:4 power splitter, both designs simulated in CST studio suit and results for S parameters presented. The designed structure is simulated in frequency range of 7GHz to 13GHz. In this design the return loss found to be less than -15dB and also shows acceptable results for transmission loss in both the power splitters, which is desirable for power splitter. RF & Microwave 2 Hall A 7/28/21 17:50
A Miniaturized Interdigital Bandpass Filter for Intentional Electromagnetic Interference Applications Abirami B (IIITDM Kancheepuram, India); Prerna Saxena (Indian Institute of Information Technology, Design and Manufacturing, Kancheepuram, India); Premkumar K. (Indian Institute of Information Technology, Design and Manufacturing Kancheepuram, India) We propose a compact bandpass filter over 0.52-2.07 GHz at the receiver front-end of an intentional electromagnetic interference detection system. We design the filter using five microstrip stubs arranged in interdigital configuration along with a dumbbell shaped defected ground structure. We design the proposed filter on FR4 substrate with a relative permittivity of 4.4 and a thickness of 1.6 mm. We obtain a 3dB fractional bandwidth of 119.96%. The proposed filter exhibits a return loss >12 dB, an insertion loss ~0.922 dB and a flat group delay over the entire bandwidth. Also, the proposed filter is compact and occupies an area of 3.55 x 1.557 cm2. As compared to the state-of-the-art designs, the proposed bandpass filter is highly miniaturized, easy to fabricate and exhibits good performance. RF & Microwave 2 Hall A 7/28/21 18:10
Performance Analysis of RIS Assisted Smart Grid HEMS Using RQAM Modulation Ashish Kumar Padhan and Pravas Ranjan Sahu (Indian Institute of Technology Bhubaneswar, India); Subhransu Samantaray (Indian Institute of Technology, Bhubaneswar, India) In this work, we analyze the performance of a reconfigurable intelligent surface (RIS) assisted radio frequency (RF) system in smart grid application. In a smart grid, the smart meter (SM) plays an important role in communication between the smart devices and the utility control centre (UCC). The UCC can communicate using the RIS assisted communication link to the SM and the SM interact with the smart devices with the communication link based RF communication using the RQAM modulation scheme. Based on the system model, a closed-form expression for the average symbol error rate (ASER) is derived and analyzed by varying the various parameters like the number of reflector in the RIS, traffic intensity, quadrature to in-phase decision distance ratio, and the total number of devices. RIS Auditorium 7/30/21 16:40
BER Analysis of RIS Assisted Bidirectional Relay System with Physical Layer Network Coding Jeba Triphena, Vetrivel Chelian Thirumavalavan and Thiruvengadam S Jayaraman (Thiagarajar College of Engineering, Madurai, India) Reconfigurable Intelligent Surface (RIS) is one of the latest technologies in bringing a certain amount of control to the rather unpredictable and uncontrollable wireless channel. In this paper, RIS is introduced in a bidirectional system with two source nodes and a Decode and Forward (DF) relay node. It is assumed that there is no direct path between the source nodes. The relay node receives information from source nodes simultaneously. The Physical Layer Network Coding (PLNC) is applied at the relay node to assist in the exchange of information between the source nodes. Analytical expressions are derived for the average probability of errors at the source nodes and relay node of the proposed RIS-assisted bidirectional relay system. The Bit Error Rate (BER) performance is analyzed using both simulation and analytical forms. It is observed that RIS-assisted PLNC based bidirectional relay system performs better than the conventional PLNC based bidirectional system. RIS Auditorium 7/30/21 17:00
Order Statistics Based Collision Analysis for PUFs Girish Vaidya, Chandramani Singh and Prabhakar Venkata T. (Indian Institute of Science, India) Physically unclonable functions (PUFs) exploit the inherent manufacturing variations for generating a device identifier. However, different devices may map to the same identifier causing a "collision". It is imperative to determine the probability of such collisions before a PUF is deployed for an application. We present a framework that computes the collision probabilities based on its inter-device and intra-device variations. This framework could be used for determining the collision probabilities, tuning the PUF attributes as well as to compare different PUF implementations. We demonstrate the use of our framework for real-world applications by comparing the results from our analyses with data from experiments and numerical simulation. Security Hall A 7/30/21 16:40
On Physical Layer Security of Correlated Multiantenna Cognitive Radio Receivers Brijesh Soni (School of Engineering and Applied Science, Ahmedabad University, India); Dhaval Karshanbhai Patel (School of Engineering and Applied Science-Ahmedabad University, India); Sagar Kavaiya (School of Engineering and Applied Science, Ahmedaad University, India); Mazen Omar Hasna (Qatar University, Qatar); Miguel López-Benítez (University of Liverpool, United Kingdom (Great Britain)) In limited space scenarios, the antennas in the multiantenna cognitive radio (CR) system are closely spaced and often experience correlation among them. In this work, the secrecy performance of correlated multiantenna CR receiver over Nakagami-m fading channels with imperfect channel state information is studied and analyzed. We consider the underlay CR paradigm wherein Alice in the secondary network communicates with Bob while Eve tries to overhear the communication. We also consider that the antennas at Bob and Eve are closely spaced and thus uniformly correlated. To this extent, we derive the analytical expressions for the first and second order secrecy measures like secrecy outage probability, average secrecy outage rate & average secrecy outage duration, respectively for the CR receiver. Moreover, in order to gain insights at high SNR, asymptotic analysis of secrecy outage probability is derived, thus obtaining the secrecy diversity gain of order L_D (i.e., the number of antennas at Bob). Monte Carlo simulations are carried out to validate the proposed analytical framework. Security Hall A 7/30/21 17:00
Detection of Speech Overlapped with Low-Energy Music Using Pyknograms Mrinmoy Bhattacharjee (Indian Institute of Technology Guwahati, India); Mahadeva Prasanna (IIT Dharwad, India); Prithwijit Guha (IIT Guwahati, India) Detection of speech overlapped with music is a challenging task. This work deals with discriminating clean speech from speech overlapped with low-energy music. The overlapped signals are generated synthetically. An enhanced spectrogram representation called Pyknogram has been explored for the current task. Pyknograms have been previously used in overlapped speech detection. The classification is performed using a neural network that is designed with only convolutional layers. The performance of Pyknograms at various high SNR levels is compared with that of discrete fourier transform based spectrograms. The classification system is benchmarked on three publicly available datasets, viz., GTZAN, Scheirer-slaney and MUSAN. The Pyknogram representation with the fully convolutional classifier performs well, both individually and in combination with spectrograms. Speech Processing 1 Hall B 7/28/21 12:10
A Spectral Variation Function for Variable Time-Scale Modification of Speech Pramod Haribhau Kachare (IIT Bombay & Ramrao Adik Institute of Technology, India); Prem C. Pandey (IIT Bombay, India) Spectral variation function is used to detect salient segments (segments with sharp spectral transitions). It is calculated from cosine of the angle between the averaged feature vectors of the adjacent segments. A modified version of this function is presented for variable time-scale modification of the speech signal. It uses the magnitude spectrum smoothed by auditory critical band filters and a small offset in the normalization for the angle cosine. Test results showed that the modified function detects spectral saliencies and does not have spurious peaks. It is applied for variable time-scale modification without altering the overall duration. Listening tests showed significantly better speech quality for processing using the modified function. Speech Processing 1 Hall B 7/28/21 12:29
Effect of High-Energy Voiced Speech Segments and Speaker Gender on Shouted Speech Detection Shikha Baghel (Indian Institute of Technology, Guwahati, India); Mahadeva Prasanna (IIT Dharwad, India); Prithwijit Guha (IIT Guwahati, India) Shouted speech detection is an essential preprocessing task in many conventional speech processing systems. Mostly, shouted speech has been studied in terms of the characterization of vocal tract and excitation source features. Previous works have also established the significance of voiced segments in shouted speech detection. This work posits that a significant emphasis is given to a portion of the voiced segments during shouted speech production. These emphasized voiced regions have significant energy. This work analyzes the effect of high-energy voiced segments on shouted speech detection. Moreover, fundamental frequency is a crucial characteristic of both shouted speech and speaker gender. Authors believe that gender has a significant effect on shouted speech detection. Therefore, the present work also studies the impact of gender on the current task. The classification between normal and shouted speech is performed using a DNN based classifier. A statistical significance test of the features extracted from high-energy voiced segments is also performed. The results support the claim that high-energy voiced segments carry highly discriminating information. Additionally, classification results of gender experiments show that gender has a notable effect on shouted speech detection. Speech Processing 1 Hall B 7/28/21 12:50
DNN Based Phrase Boundary Detection Using Knowledge-Based Features and Feature Representations from CNN Pavan Kumar J (Indian Institute of Science, India); Chiranjeevi Yarra (International Institute of Information Technology, India); Prasanta Kumar Ghosh (Indian Institute of Science, India) Automatic phrase boundary detection could be useful in applications, including computer-assisted pronunciation tutoring, spoken language understanding, and automatic speech recognition. In this work, we consider the problem of phrase boundary detection on English utterances spoken by native American speakers. Most of the existing works on boundary detection use either knowledge-based features or representations learnt from a convolutional neural network (CNN) based architecture, considering word segments. However, we hypothesize that combining knowledge-based features and learned representations could improve the boundary detection task's performance. For this, we consider a fusion-based model considering deep neural network (DNN) and CNN, where CNNs are used for learning representations and DNN is used to combine knowledge-based features and learned representations. Further, unlike existing data-driven methods, we consider two CNNs for learning representation, one for word segments and another for word-final syllable segments. Experiments on Boston University radio news and Switchboard corpora show the benefit of the proposed fusion-based approach compared to a baseline using knowledge-based features only and another baseline using feature representations from CNN only. Speech Processing 1 Hall B 7/28/21 13:10
Towards a Database for Detection of Multiple Speech Disfluencies in Indian English Sparsh Garg and Utkarsh Mehrotra (International Institute of Information Technology Hyderabad, India); Krishna Gurugubelli (IIIT-Hyderabad, India); Anil Kumar Vuppala (International Institute of Information Technology Hyderabad, India) The detection and removal of disfluencies from speech is an important task since the presence of disfluencies can adversely affect the performance of speech-based applications such as Automatic Speech Recognition (ASR) systems and speech-to-speech translation systems. From the perspective of Indian languages, there is a lack of studies pertaining to speech disfluencies, their types and frequency of occurrence. Also, the resources available to perform such studies in an Indian context are limited. Through this paper, we attempt to address this issue by introducing the Indian English Disfluency (IED) Dataset. This dataset consists of 10-hours of lecture mode speech in Indian English. Five types of disfluencies - filled pause, prolongation, word repetition, part-word repetition and phrase repetition were identified in the speech signal and annotated in the corresponding transcription to prepare this dataset. The IED dataset was then used to develop frame-level automatic disfluency detection systems. Two sets of features were extracted from the speech signal and then used to train classifiers for the task of disfluency detection. Amongst all the systems employed, Random Forest with MFCC features resulted in the highest average accuracy of 89.61% and F1-score of 0.89. Speech Processing 1 Hall B 7/28/21 13:30
Instantaneous Frequency Filter-Bank Features for Low Resource Speech Recognition Using Deep Recurrent Architectures Shekhar Nayak (Samsung R&D Institute Bangalore & IIT Hyderabad, India); Chintigari Shiva Kumar (IIT Hyderabad, India); Sri Rama Murty Kodukula (Indian Institute of Technology Hyderabad, India) Recurrent neural networks (RNNs) and its variants have achieved significant success in speech recognition. Long short term memory (LSTM) and gated recurrent units (GRUs) are the two most popular variants which overcome the vanishing gradient problem of RNNs and also learn effectively long term dependencies. Light gated recurrent units (Li-GRUs) are more compact versions of standard GRUs. Li-GRUs have been shown to provide better recognition accuracy with significantly faster training. These different RNN inspired architectures invariably use magnitude based features and the phase information is generally ignored. We propose to incorporate the features derived from the analytic phase of the speech signals for speech recognition using these RNN variants. Instantaneous frequency filter-bank (IFFB) features derived from Fourier transform relations performed at par with the standard MFCC features for recurrent units based acoustic models despite being derived from phase information only. Different system combinations of IFFB features with the magnitude based features provided lowest PER of 12.9% and showed relative improvements of up to 16.8% over standalone MFCC features on TIMIT phone recognition using Li-GRU based architecture. IFFB features significantly outperformed the modified group delay coefficients (MGDC) features in all our experiments. Speech Processing 2 Hall B 7/28/21 14:30
CTC-Based End-To-End ASR for the Low Resource Sanskrit Language with Spectrogram Augmentation Anoop Chandran Savithri (Indian Institute of Science, Bangalore, India); Ramakrishnan Angarai Ganesan (Indian Institute of Science & RaGaVeRa Indic Technologies Pvt Ltd, India) Sanskrit is one of the Indian languages which fares poorly, with regard to the development of language based tools. In this work we build a connectionist temporal classification (CTC) based end-to-end large vocabulary continuous speech recognition system for Sanskrit. To our knowledge, this is the first time an end-to-end framework is being used for automatic speech recognition in Sanskrit. A Sanskrit speech corpus with around 5.5 hours of speech data is used for training a neural network with CTC objective. 80-dimensional mel-spectrogram together with their delta and delta-delta are used as the input features. Spectrogram augmentation techniques are used to effectively increase the amount of training data. The trained CTC acoustic model is assessed in terms of character error rate (CER) on greedy decoding. Weighted finite state transducer (WFST) decoding is used to obtain the word level transcriptions from the character level probability distributions obtained at the output of the CTC network. The decoder WFST, which maps the CTC output characters to the words in the lexicon, is constructed by composing 3 individual finite state transducers (FST), namely token, lexicon and grammar. Trigram models trained from a text corpus of 262338 sentences are used for language modeling in grammar FST. The system achieves a word error rate (WER) of 7.64% and a sentence error rate (SER) of 32.44% on the Sanskrit test set of 558 utterances with spectrogram augmentation and WFST decoding. Spectrogram augmentation provides an absolute improvement of 13.86% in WER. Speech Processing 2 Hall B 7/28/21 14:50
Spoken Language Diarization Using an Attention Based Neural Network Jagabandhu Mishra and Ayush Agarwal (Indian Institute of Technology Dharwad, India); Mahadeva Prasanna (IIT Dharwad, India) Spoken language diarization (SLD) is a task to perform automatic segmentation and labeling of the languages present in a given code-switched speech utterance. Inspiring from the way humans perform SLD (i.e capturing the language specific long term information), this work has proposed an acoustic-phonetic approach to perform SLD. This acoustic-phonetic approach consists of an attention based neural network modelling to capture the language specific information and a Gaussian smoothing approach to locate the language change points. From the experimental study, it has been observed that the proposed approach performs better when dealing with code-switched segment containing monolingual segments of longer duration. However, the performance of the approach decreases with decrease in the monolingual segment duration. This issue poses a challenge in the further exploration of the proposed approach. Speech Processing 2 Hall B 7/28/21 15:10
Speech-Training Aid with Time-Scaled Audiovisual Feedback of Articulatory Efforts Pramod Haribhau Kachare (IIT Bombay & Ramrao Adik Institute of Technology, India); Prem C. Pandey (IIT Bombay, India); Vishal Mane (India, India); Hirak Dasgupta (IIT Bombay, India); K. S. Nataraj (Indian Institute of Technology, Bombay, India) Hearing-impaired children lack auditory feedback and experience difficulty in acquiring speech production. They can benefit from speech training aids providing visual feedback of key articulatory efforts. Requirements for such aid are developed through extended interaction with speech therapists and special education teachers. The aid is developed as a PC-based app for ease of distribution and use. It has two panels to enable comparison between the articulatory efforts of the learner and a teacher or a pre-recorded reference speaker. The visual feedback for an utterance is based on the information obtained from its audiovisual recording. The speech signal is processed to obtain time-varying vocal tract shape, level, and pitch. The vocal tract shape estimation uses LP-based inverse filtering, and the pitch estimation uses glottal epoch detection using Hilbert envelope for excitation enhancement. Visual feedback comprises a variable-rate animation of the lateral vocal tract shape, level, and pitch, and time-aligned display of the frontal view of the speaker's face along with playback of time-scaled speech signal. The graphical user interface and modules for signal acquisition, speech analysis, and time-scaled animation are developed and integrated using Python. The app has been tested for its functionalities and user interface and needs to be evaluated for speech training of hearing-impaired children. It may also be useful to second-language learners in improving the pronunciation of unfamiliar sounds. Speech Processing 2 Hall B 7/28/21 15:30
Joint Bandwidth and Position Optimization in UAV Networks Deployed for Disaster Scenarios Neetu R R (IIIT Delhi, India); Akshita Gupta (Indraprastha Institute of Information Technology, India); Gourab Ghatak (IIIT Delhi, India); Anand Srivastava (Indraprastha Institute of Information Technology Delhi, India); Vivek A Bohara (Indraprastha Institute of Information Technology, Delhi (IIIT-Delhi), India) In this paper, we consider a disaster-affected area that has lost cellular connectivity. In such a scenario, we study a unmanned aerial vehicle (UAV) network deployed to restore the user connections to the core network. In particular, we propose a resource partitioning scheme for integrated access and backhaul in the downlink to sustain the user connections. First, we derive an analytical expression for the throughput coverage probability for the access (i.e., UAV to user) and the backhaul (i.e., terrestrial base station (BS) to UAV) links. Then, we investigate the optimal position of the UAV and the optimal partitioning of the frequency resources between the access and the backhaul links, in order to maximize the coverage of the users. Our study highlights that as the number of users in the network and/or their throughput demand increases, the optimal UAV position moves towards the centre of the disaster-affected area. Moreover, in that case, a higher fraction of the available frequency resources must be provisioned for the access links. On the contrary, when the backhaul throughput requirements are high, or in the case of sparsely deployed terrestrial BSs, the optimal UAV position is at the edge of the disaster-affected area. Thus, this study provides key system-design insights to a network operator for deploying emergency areal networks to extend the cellular coverage. UAV Communications Hall A 7/30/21 14:00
Trajectory Prediction of UAVs for Relay-Assisted D2D Communication Using Machine Learning Pradip Kumar Barik (Indian Institute of Technology, Kharagpur, India); Ashu Dayal Chaurasiya, Raja Datta and Chetna Singhal (Indian Institute of Technology Kharagpur, India) Device-to-Device (D2D) communication has been proven an efficient technique in the present and upcoming cellular networks for improving network performance. Many a time, a direct D2D link may not be available due to longer distance or poor channel quality between two devices. Multi-hop D2D is an effective solution to overcome this limitation of direct D2D communication. Here relay devices help in forwarding data from transmitters to the receivers through single or multiple hops. However, finding suitable fixed relays and their locations is a complex problem, which does not have an efficient solution. In this paper, we have used UAVs (drones) that act as relays for forwarding data between two devices. The proposed approach serves more out of direct range D2D users resulting in a reduced churn rate of the system. We find the trajectory of such UAVs with the help of active user prediction using Neural Networks (NN) to serve all the D2D users by increasing the coverage range of D2D communications. We have estimated the number of active D2D users in every zone covered by each drone and intra and inter-drone communication trajectories. It is also shown that the packet loss ratio remains within the acceptable limit for the proposed trajectories of the UAVs by choosing a sufficient buffer length. UAV Communications Hall A 7/30/21 14:20
Optimal Deployment Strategy for Relay Based UAV Assisted Cooperative Communication for Emergency Applications Nelapati Lava Prasad, Chanakya Ajit Ekbote and Barathram. Ramkumar (Indian Institute of Technology Bhubaneswar, India) There has been a lot of research in the last decade on UAV assisted wireless communication networks. Due to its ability of fast deployment, it is seen as a potential solution to establish communication under emergency scenarios like natural disasters. The mobile nature of the UAVs offers a lot of flexibility, which can be harnessed to improve the QoS of a wireless communication network. In this paper UAV assisted cooperative communication to serve different user clusters distributed in a geographical location is considered. These user clusters do not have access to any conventional base station which is typically a scenario under natural disasters. Each cluster is served by two types of UAVs: cluster UAV which hovers on the top of the cluster centroid and relay UAV which relays information between a central base station (CBS) and cluster UAV. To achieve the required QoS, which is serving a maximum number of users with limited available power, two major parameters have to be optimized apart from other parameters. These are the height of the cluster UAV and trajectory of the relay UAV. To solve this problem, a three-step approach is considered in this paper. In the first step, an unsupervised learning algorithm is used to find the horizontal location parameters of cluster UAVs. Then using convex optimization to find the optimal height of the cluster UAV under power constraints and capacity requirement. Finally using a heuristic algorithm to find the optimal trajectory with minimum distance to be traveled by the relay UAVs. The wireless channel considered here is a simple line of sight (LoS) with a path loss. Computer simulations are performed to prove the validity of the proposed approach in comparison with random deployment. UAV Communications Hall A 7/30/21 14:40
Optimal Data Transfer in UAV-Assisted Edge-Networks Using 3D Beamforming Shraddha Tripathi (Indian Institute of Technology Kanpur, India & NCTU Taiwan, Taiwan); Om Jee Pandey (University of Saskatchewan, Canada); Rajesh M Hegde (Indian Institute of Technology Kanpur, India) Reliable and low-latency data transfer to the cell edge users (CEUs) of 5G edge-network is a challenging problem. Solution to this problem can enable real-time applications such as remote health-monitoring of patients and target tracking in battle field. In this work, a novel method for optimal data transfer over UAV-assisted edge-networks is proposed. The proposed method utilizes unmanned aerial vehicle (UAV) as a relay node for data transfer between ground base station (GBS) and the CEUs. Additionally, UAV node is designed to be able to perform 3D beamforming leading to improved signal to interference noise ratio (SINR) and high throughput. To obtain optimal data transfer, the CEUs are first geographically clustered using a distance criterion. Subsequently, a joint optimization problem that aims to find the UAV trajectory and the beamforming downtilt angles, while applying minimum latency and maximum throughput constraints is formulated. This joint optimization problem is solved by using an iterative approach. Extensive simulations are then performed to validate this method for network latency and throughput under varying network conditions. The results are motivating enough for the method to be used in medium and large scale edge networks. UAV Communications Hall A 7/30/21 15:00
DeepSCT: Deep Learning Based Self Correcting Object Tracking Mechanism Khush Agrawal, Rohit Lal and Himanshu Patil (Visvesvaraya National Institute of Technology, India); Surender Kannaiyan (VNIT, Nagpur, India); Deep Gupta (Visvesvaraya National Institute of Technology, Nagpur, India) This paper presents a novel mechanism, DeepSCT, to handle the long-term object tracking problem in Computer Vision. The paper builds around the premise that the classical tracking algorithms can handle short-term tracking problems efficiently; however, they fail in the case of long-term tracking due to several environmental disturbances like occlusion and out-of-frame going targets. The relatively newer Deep Learning based trackers have higher efficacy but suffer from working in real-time on low-end hardware. We try to fuse the two methods in a unique way such that the resulting algorithm has higher efficiency and accuracy simultaneously. We present a modular mechanism, which can accommodate improvements in its sub-blocks. The algorithm was tested on the VisDrone-SOT2019 dataset for a person tracking task. We quantitatively and qualitatively show that DeepSCT significantly improved classical algorithms' performance in short-term and long-term tracking problems. Video Processing 1 Hall B 7/28/21 16:00
Domain Randomization on Deep Learning Models for Image Dehazing Abdul Fathaah Shamsuddin, Abhijith P, Deepak Raja Sekar P M, Krupasankari Ragunathan and Praveen Sankaran (National Institute of Technology Calicut, India) Haze is a naturally occurring phenomenon that obstructs vision and affects the quality of images and videos. Recent literature has shown that deep learning-based image dehazing gives promising results both in terms of image quality and execution time. However, the difficulty of acquiring real-world hazy -- clear paired images for training still remains a challenge. Widely available datasets use synthetically generated hazy images that suffer from flaws due to difficulty in acquiring accurate depth information to synthesize realistic-looking haze, causing a gap in the real and synthetic domain. In this paper, we propose the usage of domain randomization for image dehazing by generating a completely simulated training dataset for deep learning models. A standard UNET based dehazing model is trained on the simulated dataset without using any real-world data to obtain high quality dehazed images. The performance of the proposed approach is evaluated on the Sun-Dehaze dataset and RESIDE Standard (SOTS outdoor) dataset. We obtain favorable PSNR and SSIM scores on both sets and we also show how our approach yields better visual results compared to other learning-based approaches. Video Processing 1 Hall B 7/28/21 16:20
Deep Video Compression Using Compressed P-Frame Resampling Abhishek Kumar Sinha (Indian Institute of Space Science and Technology, India); Deepak Mishra (IIST, India) The Convolutional Neural Network has emerged as one of the major players in the field of deep video compression. Many deep learning models relying on convolutional layers have outperformed the state-of-the-art compression standards by a huge margin. Although their work is still at infancy level, they seem to be the future of video coding. The proposed approach uses a frame resampling based video compression approach using Temporal 3-D CNN based encoder and Y-style CNN based decoder concatenated with High Fidelity GAN based entropy coding for frame compression. The proposed architecture employs frame downsampling method over the residual frame to control the bitrate of the compressed data and is trained through a simplified stagewise training procedure. The extensive experiments are conducted with different datasets and different colorspaces. The study shows that the proposed model outperforms the H.265 by 0.255 dB in terms of PSNR and nearly 0.1 in terms of MS-SSIM. Video Processing 1 Hall B 7/28/21 16:40
Forensics of Decompressed JPEG Color Images Based on Chroma Subsampling Chothmal Kumawat (IIT Roorkee, India); Vinod Pankajakshan (Indian Institute of Technology Roorkee, India) Identification of the type of chroma subsampling in a decompressed JPEG color image stored in a lossless format is important in forensic analysis. It is useful in many forensic scenarios like detecting localized forgery and estimating the quantization step sizes in the chroma planes for source camera identification. In this work, we propose a machine learning-based method capable of identifying the chroma subsampling used in the compression process. The method is based on detecting the change in adjacent pixel correlations due to upsampling process in JPEG decompression. These changes in the correlation are measured using the two-sample KS-test statistic in different directions. The experimental results show the efficacy of the proposed method in identifying the chroma subsampling scheme. Video Processing 1 Hall B 7/28/21 17:00
An Optical Flow Based Approach to Detect Movement Epenthesis in Continuous Fingerspelling of Sign Language Navneet Nayan (Indian Institute of Technology, Roorkee, India); Pyari Mohan Pradhan (IIT Roorkee, India); Debashis Ghosh (Indian Institute of Technology (IIT) Roorkee, India) In this paper, the movement epenthesis problem in continuous fingerspelling is addressed. Movement epenthesis caused due to unwanted but unavoidable hand movement in between two sign gestures in continuous signing is one of the major problems in automatic sign language recognition. A novel method based on calculating the 2-norm values of the magnitude matrices of optical flow has been proposed in this paper to detect the movement epenthesis containing video frames. We used Horn-Schunck method to compute the optical flow and estimate the speed of hands in the the continuous fingerspelling videos. The 2-norm values of the magnitude matrix of optical flow provides a discriminative feature to distinguish of movement epenthesis frames from sign frames and hence mean of the 2-norm values is used as the threshold value to detect the movement epenthesis frames in a gesture video. We tested our method on continuous fingerspelling videos of Indian sign language. Experimental results show that the performance of our proposed method is 100% accurate in detecting the movement epenthesis frames in continuous fingerspelling. Video Processing 2 Hall B 7/28/21 17:30
A Level Set Model Driven by New Signed Pressure Force Function for Image Segmentation Soumen Biswas (National Institute of Technology Silchar, India); Ranjay Hazra (Nit Silchar, India); Shitala Prasad (Research Scientist, Singapore); Arvind Sirvee (Rajasthan Technical University, Kota, India) An image segmentation model using histogram-based image fitting (HF) energy is proposed to identify objects with poorly defined boundaries. The proposed energy model considers an improved fitting energy function based on normalized histogram and average intensities of objects inside as well as outside the contour curve. The fitting energy functions are computed before the curve evolution thereby reducing the complexity of intensity inhomogeneity images. Further, a new signed pressure force function is incorporated in the proposed energy model which can increase the efficiency of the curve evolution process at blur edges or at weak edge regions. The comparative analysis of the proposed energy model produces better segmentation results compared to the other state-of-the-art energy models namely the Li et. al. model, local binary fitting (LBF), and Chen-Vese (C-V) models. The proposed model is also robust to intensity inhomogeneity. In addition, the calculation of the Jaccard Index (JI) proves the robustness of the proposed energy model. Video Processing 2 Hall B 7/28/21 17:50
Phrase Recognition Using Improved Lip Reading Through Phase-Based Eulerian Video Magnification Salam Nandakishor and Debadatta Pati (National Institute of Technology Nagaland, India) Lip reading is a technique to understand speech by visual observations of the lip movements. While speaking the subtle motion or temporal variations of our mouth are generally invisible by naked humans eyes. It is mainly due to the limited range of visual perception. These imperceptible visual information consist of useful hidden information. The Eulerian video magnification (EVM) technique is used to magnify the video for revealing such hidden information. In this work, the phase based EVM method is used to magnify the subtle spatial and temporal information of the mouth movements for phrases recognition task. The local binary pattern histogram extracted from three orthogonal plane (XY, XT and YT), known as LBP-TOP is used as visual feature to represent mouth movements. The support vector machine (SVM) is used for recognition of phrases. The experiments are performed on OuluVS database. The lip-reading approach without EVM provides 62% accuracy whereas the phase based EVM method provides 70% accuracy. This shows that the proposed method extracts comparatively more robust and discriminative visual features for phrase recognition task. Video Processing 2 Hall B 7/28/21 18:10
Visibility Restoration of Diverse Turbid Underwater Images- Two Step Approach Mary S Cecilia (Anna University & SSN College of Engineering, India); Sakthivel Murugan Santhanam (SSN College of Engineering, India) Underwater Images are of degraded quality due to the scattering and absorption. The color cast and turbidity that hinder the visibility of such images are due to the sediments present that vary for diverse environments. Shallow water images are very turbid. The images too suffer from negative effects of artificial illumination when capturing data. Here a two-step approach is formulated to restore and enhance the underwater images from different locations. The images are then blended using a wavelet fusion considering the mean of the images. The output images demonstrate reduced haze, improved contrast and enhanced sharpness with adequate removal of the color cast. The results project better visibility on both subjective and objective measures compared to recent restoration and enhancement methods. Video Processing 2 Hall B 7/28/21 18:30
Optimal Link Scheduling for Low Latency Data Transfer over Small World WSNs Om Jee Pandey (University of Saskatchewan, Canada); Naga Srinivasarao Chilamkurthy (University of SRM AP, India); Rajesh M Hegde (Indian Institute of Technology Kanpur, India) In recent years, small world characteristics (SWC) received huge attention due to their various advantages in the context of social, electrical, computer, and wireless networks. A wireless sensor network (WSN) exhibiting SWC is known as small world WSN (SW-WSN). Therefore, SW-WSN consists small average path length and large average clustering coefficient. Here, in this paper, a novel optimal link scheduling method is proposed to develop SW-WSN. The proposed method determines, optimal number of new links need to be created in the network. Additionally, the method also finds the optimal node-pairs towards creation of these links. The developed algorithm considers node betweenness centrality measure for the introduction of SWC. SW-WSN obtained using proposed method yields reduced time complexity towards its development. Moreover, it also results in optimal SWC when compared to other existing methods. A reduced data transmission delay is noted over SW-WSN developed using proposed method. Random, near-optimal, and sub-optimal methods of introducing SWC and their time complexities are also investigated and compared to the proposed method. The results are computed over simulated and real WSN testbed. Obtained results demonstrate the significance of proposed method and its utilization over large scale network applications. WSN Hall A 7/29/21 8:20
Efficient Message Dissemination in V2V Network: A Local Centrality-Based Approach Moyukh Laha (IIT Kharagpur, India); Raja Datta (Indian Institute of Technology Kharagpur, India) Many vehicular applications require data dissemination where all the vehicles in a specific region of concern are the intended receivers of particular messages. Such dissemination is challenging due to vehicular networks' distinct properties, such as high mobility, low communication range, intermittent connectivity, and diverse variations in their topology. In this work, we propose a Local Centrality-based Dissemination scheme for vehicular networks based on V2V communication. To this end, each vehicle node gathers their two-hop neighborhood information to identify the super-spreader nodes that continue the dissemination by rebroadcasting the receiving messages. In contrast, the rest of the nodes remain quiet. We validate the performance of our proposed scheme with real vehicular data. Extensive simulation results reveal the superior performance of our proposed scheme in terms of higher and quicker coverage with fewer redundant transmissions than the state-of-the-art data dissemination protocols. WSN Hall A 7/29/21 8:40