Muralishankar R
@ewit.edu.in
Professor, Department of Electronics and Communication Engineering
Dean - Academics & Research, East West Institute of Technology, Bengaluru 560091
EDUCATION
2003
Ph. D.
Indian Institute of Science, Bangalore
RESEARCH, TEACHING, or OTHER INTERESTS
Engineering, Electrical and Electronic Engineering, Biomedical Engineering, Signal Processing
Scopus Publications
Scopus Publications
Sutapa Sarkar, R. Muralishankar, and Sanjeev Gurugopinath
Institution of Engineering and Technology (IET)
AbstractAccurate detection of spectrum holes is a useful requirement for cognitive radios that improves the efficiency of spectrum usage. The authors propose three novel, simple, and entropy‐based detectors for spectrum sensing in cognitive radio. The authors evaluate the probability of detection of these three detectors: Vasicek's entropy detector, truncated Vasicek's entropy detector, and Van Es' entropy detector, over a predefined probability of false‐alarm. In particular, the authors provide the approximate and asymptotic test statistics for these detectors in the presence and absence of Nakagami‐m fading, noise variance uncertainty, and optimised detection threshold. Furthermore, the authors provide a detailed comparison study among all the detectors via Monte Carlo simulations and justify authors results through real‐world data. The authors’ experimental results establish a superior performance of truncated Vasicek's entropy detector over Vasicek's entropy detector, energy detector, differential entropy detector and Van Es' entropy detector in practically viable scenarios.
Arati Halaki, Sutapa Sarkar, Sanjeev Gurugopinath, and R. Muralishankar
Institution of Engineering and Technology (IET)
AbstractCognitive radio (CR) systems are configured to dynamically assess the spectrum utilisation and contribute towards an improved spectrum efficiency. Hence, accurate detection of the incumbent signal in a given channel, popularly known as spectrum sensing (SS), is essential for CR. Here, in the domain of SS, the authors introduce a new goodness‐of‐fit test (GoFT) founded on p‐norm of the observations at the receiver node. To capture the heavy‐tailed nature of noise distribution in practical communication channels, the authors utilise generalised Gaussian distribution (GGD) as a noise model. A novel p‐norm detector (PND) and a geometric power detector (GPD) is proposed and corresponding probability density function (PDF) under GGD is derived. Via Monte Carlo simulations, the authors show a match of the derived PDFs with the simulation results. Using Neyman‐Pearson framework the performances of PND and GPD are compared with an existing differential entropy detector (DED), the well‐known energy detector (ED) and joint correlation and energy detector (CED) under GGD noise model. Evaluation of proposed PND and GPD utilising Monte Carlo simulations indicate a superior performance. Further, the experiments employing real‐world data establish superiority of the proposed detectors as compared to existing techniques. The authors derive and implement an optimised threshold for PND, providing further improvement in performance.
Roopesh Ramesh, R. Muralishankar, and Sanjeev Gurugopinath
IEEE
In this paper, we study and present the performance based on the outage probability of a single-stage, relay-aided, cooperative non-orthogonal multiple access over power line communication under a Bernoulli-Gaussian impulsive noise. The network setup comprises one source node and two destination nodes, among which one of the nodes closer to the source acts as a relay, transmitting a symbol from the source node to the destination. The relay considered is a decoded-and-forward relay, which helps to transmit the source symbol to the destination over two-time frames. We derive the mathematical expressions to determine outage probabilities at the relay and destination nodes within the studied network. Furthermore, we develop an optimization problem related to the power distribution coefficient at the source, deriving the corresponding mathematical expressions. To validate our analysis, we employ simulations and numerical methods.
Roopesh Ramesh, R. Muralishankar, and Sanjeev Gurugopinath
IEEE
This paper studies the performance of outage probability in a dual-stage (DS) cooperative non-orthogonal multiple access (NOMA) system for power line communication (PLC) in the presence of Bernoulli-Gaussian impulsive noise. The network configuration includes a single source node and two destination nodes among which one the node act as a decode-and-forward (DF) relay node. This is similar to the existing research on DS NOMA setup, which primarily focused on additive white Gaussian noise. In this setup, both the source and relay transmit data to the destination node using NOMA across two consecutive time frames. We provide mathematical expressions for outage probabilities at both nodes in the context of Bernoulli-Gaussian noise. Furthermore, our analysis is validated using numerical methods and Monte Carlo simulations. Additionally, our results indicate that a DS-NOMA configuration exhibits lower outage probability when compared to a single-stage NOMA setup.
R. Muralishankar, Debayan Ghosh, and Sanjeev Gurugopinath
Elsevier BV
Sutapa Sarkar, R. Muralishankar, and Sanjeev Gurugopinath
IEEE
In this paper, we propose an active threshold correction (ATC) for the generalized energy detector (GED) under spectrum sensing to alleviate the effects of noise variance uncertainty (NVU) and fading. First, we derive the probability of false alarm and the probability of detection for GED, with an arbitrary exponent, assuming that the test statistic follows a gamma distribution and under Nakagami-m fading. Next, we present the effect of NVU on the performance of GED. Later, we derive an analytic expression for the corresponding ATC under NVU and Nakagami-m fading. Through numerical method, we show that by employing the ATC, the performance of GED improves in the presence of NVU and fading. Moreover, in the considered setup for GED, we show that the GED outperforms the energy detector.
Shreeram Suresh Chandra, Akshay Upadhye, Purushothaman Saravanan, Sanjeev Gurugopinath, and R. Muralishankar
IEEE
In this work, we consider a performance comparison of deep learning-based approaches to the problem of spectrum sensing (SS) in cognitive radios. Towards this end, we use signal processing (SP) features such as energy, differential entropy, geometric power and p-norm. For the classification problem of SS, we employ deep neural network (NN) architectures such as multi-layer perceptron (MLP), convolutional NN, fully convolutional network, residual NN (ResNet), long short-term memory and temporal convolutional network. Through extensive experiments based on real-world captured datasets and Monte Carlo simulations, we show that MLP and ResNet architectures offer the best performance in terms of probability of detection, for a given predefined level of probability of false-alarm. Further, we show that NN architectures trained with a combined set of the SP features yield the best performance.
R. Muralishankar, Debayan Ghosh, and Sanjeev Gurugopinath
Institute of Electrical and Electronics Engineers (IEEE)
We propose a novel modified Mel-discrete cosine transform (MMD) filter bank structure, which restricts the overlap of each filter response to its immediate neighbor. In contrast to the well-known triangular filters employed in the extraction of the Mel-frequency cepstral coefficients (MFCC), the proposed filter structure has a smoother response and offers discrete cosine transformation and Mel-scale filtering in a single operation. It is known that the choice of MFCC as the only feature for voice activity detection (VAD) does not yield substantial improvements in the performance. Even with the long-term approach, we observe a not so encouraging VAD performance when MFCC features are employed. However, other long-term based VAD algorithms – without MFCC – are known to provide a substantial improvement in the performance under low SNR with time-varying statistics of speech and/or noise. In this work, we show that by employing the MMD followed by the long-term differential entropy of voice signal for VAD provides significant improvements in detection accuracy when compared with the other well-known long-term algorithms. Thus, this study opens up the possible benefits of the proposed MMD filter bank for other speech processing applications.
Sutapa Sarkar, R. Muralishankar, and Sanjeev Gurugopinath
IEEE
In this paper, we employ generalized energy detector (GED) (with parameter, p) for spectrum sensing and study its performance under additive white Gaussian noise with Gaussian primary user statistics. Here, we consider the distribution of test statistics of GED to follow Gamma distribution. To evaluate the performance of GED, we setup the test bench under Gamma distribution of test statistics to obtain probability of detection over probability of false alarm and probability of error (PE) over signal-to-noise ratio (SNR). Later, we optimize the threshold for GED under Gamma distribution for test statistics and obtain a closed form expression for optimal threshold. The optimal threshold is employed to study the performance of GED. The enhanced performance of GED with optimal threshold is observed under Gamma distributed test statistics, and in particular, for GED parameter p = 1 under Gamma distribution. This choice of $p$ consistently provides least PE over SNR under optimized threshold setup. However, without employing optimized threshold, the performance of GED for different values of $p$ is almost similar. In our performance analysis, we choose the number of samples, $N = 10$ which means that the demand for sensing time is less. Further, we study the performance of the GED under the influence of different types of fading channels. Again we observe the performance of GED does not change much for different $p$ values under the influence of fading.
Sutapa Sarkar, R. Muralishankar, and Sanjeev Gurugopinath
IEEE
In this paper, we consider the generalized energy detector (GED) for spectrum sensing in cognitive radios using a Bayesian approach. First, we derive the asymptotic distribution of the GED test statistics under both hypotheses using the central limit theorem. We then obtain a closed-form solution for the optimal detection threshold that minimizes the probability of overall error – defined as the linear combination of the false-alarm probability and mis-detection probability. The parameter of the GED is also chosen to minimize the probability of error. We validate our theory through Monte Carlo simulations. Additionally, we also investigate the performance degradation of GED under noise variance uncertainty.
Arati Halaki, Sanjeev Gurugopinath, and R. Muralishankar
IEEE
Spectrum sensing is a important feature in cognitive radios, wherein the idea is to detect the presence of the incumbent on a licensed spectrum. It is known that the improved energy detector (IED)- which calculates the sample norm of an optimally chosen order in the received observations - outperforms the conventional energy detector (ED). The relatively recently proposed joint correlation and energy detector (CED) combines the sample energy and the first order correlation values in the received observations, and it is known that when the combining is carried out optimally, it outperforms ED. We present a comparative performance study of CED and IED, for different incumbent signal models, following the Neyman-Pearson framework. We show that the CED consistently outperforms IED across the Gaussian, sinusoidal and constant signal models. Similar to what was reported earlier, the IED is observed to outperform ED only for low values of probability of false-alarm. Hence, this study establishes the applicability of CED for spectrum sensing, as opposed to IED and ED.
Raveesh Hegde and R. Muralishankar
IEEE
There are many features proposed in the literature for voice activity detection (VAD). Shen et al. [20] first used a spectral entropy-based feature to detect regions of speech spurts under noisy conditions. However, VAD employing this feature was unreliable when the noise level greatly exceeds the speech level. To improve the performance of spectral entropy based VAD under low signal-to-noise ratios (SNRs), spectrum of a signal over a frame is divided into sub bands and spectral entropy is computed over these bands. Later, these spectral entropies are weighted and summed to obtain the entropy. Based on the amount of noise in each band, weights were found empirically. This approach was named as banded spectral entropy (BSE) [21]. In [24], deviation threshold computed from approximate ramp line and the sorted spectral coefficients of the band are adopted to decide useful/useless bands. In this paper, we propose a novel Teager Energy Band Spectral Entropy (TE_BSE) feature for VAD. Here, we carryout enhancement of spectral peaks employing Teager energy of each frequency transformed speech frame. This is followed with dividing of spectrum into sub bands and entropy computation over each band. The summing of entropy from each useful band is done to get TE _ BSE feature. We identify useful/useless bands following [24]. Later, we present the performance of our proposed VAD in terms of probability of detection $(\\pmb{P}_{\\pmb{D}})$, probability of false alarm $(\\pmb{P}_{\\pmb{FA}})$ and probability of error under different noises and SNRs. Finally, from the VAD results on real-world sample, proposed VAD outperforms statistical based VAD by Sohn et. al. [8] with improved $\\pmb{P}_{\\pmb{D}}$ not at the cost of increase in $\\pmb{P}_{\\pmb{FA}}$.
Roopesh R., B.S. Sushma, Sanjeev Gurugopinath, and R. Muralishankar
IEEE
We study the channel capacity of a narrowband powerline communication (NBPLC) – operating in the range of 3 - 500 kHz – channel in the presence of a simplified additive white class A impulsive noise (SAWCN). According to this model, the receiver noise in the NBPLC can be modeled as a two-term Gaussian mixture distribution, consisting of the background noise component and an impulsive component. First, we consider the practically relevant case in which the transmit power is much less than the variance of the impulsive component, where a tight approximation to the entropy of the SAWCN is calculated to derive tight lower and upper bounds on the channel capacity. Next, we consider the case where the transmit power is higher and derive tighter lower and upper bounds. Our numerical results indicate a small gap between the derived bounds, which emphasize their potential applications in serving as meaningful estimates on the achievable rate of NBPLC in the presence of impulsive noise. These bounds are closely achievable when the input signal is Gaussian.
Sujay Narayana, R. Muralishankar, R. Venkatesha Prasad, and Vijay S. Rao
ACM
Two nanosatellites recently launched into space had issues with respect to its stabilization, power and orientation. The signals were intermittent, and amateur radio enthusiasts around the globe were requested to observe the satellites so as to get their health information. As decoding the received signals required proprietary hardware (that could not be sent to everyone), amateur radio receivers recorded the signal using Software Defined Radios (SDRs) and subsampled the carrier signals to make it easy to share. The captured signals, modulated using binary Frequency Shift Keying (FSK), included noise and more importantly the frequency shifts due to Doppler, caused by the speed of the satellites (of about 7.8 km/s), thus making decoding a major challenge even for the designated proprietary receivers (failed in some cases). As the existing FSK methods did not work effectively, we were motivated by this challenge to design an effective FSK decoder that works in the presence of Doppler and noise. In this paper, we propose Teager Energy Decoder (TED) based on Teager Energy Operator to decode such Doppler and noise influenced sub-sampled data. TED does not need any Doppler correction mechanisms and can dynamically adapt to the changing frequency shifts. We evaluate TED using simulation as well as from the signals from those two satellites. We show that TED performs better than COTS transceivers and available GNU-radio-based solutions using SDRs. TED is low-complexity algorithm,O(N2), and has been prototyped on a low-power microcontroller. TED can be easily adopted on satellites to decode signals for space Internet of Things applications.
B.S. Sushma, R. Roopesh, Sanjeev Gurugopinath, and R. Muralishankar
IEEE
In this paper, we study the fundamentals of the broadband powerline communication (BBPLC) for internet-of-things (IoT) applications, operating in the range of 1.2–300 MHz. First, we consider the BBPLC channel model by considering the joint statistics of the magnitude and phase of the channel frequency response. Then, we describe the widely used noise models in BBPLC, and study the detection performance of BPSK symbols over the BBPLC channel impaired with the asynchronous impulsive noise and the simplified Middleton’s class A noise, over fading. Later, we consider the impact of the simplified Middleton’s noise model on the BBPLC channel by evaluating the normalized capacity, by modeling the channel as an irreducible, stationary, aperiodic Markov chain with two states. Through numerical results, we compare the capacity obtained for this channel with that of the typical Middleton’s class A channel, and show that the approximation due the simplified model is negligible when the ratio of the impulsive component to the background component is not very large. Moreover, the capacity derived using the proposed approach serves as a tight upper bound on the capacity of the class A channel.
Arati Halaki, C.A. Manohar, Sanjeev Gurugopinath, and R. Muralishankar
IEEE
We propose a novel spectrum sensing algorithm based on the differential entropy under symmetric alpha stable impulsive noise, in contrary to the several existing algorithms where the Gaussian noise is assumed. The need for a non-Gaussian noise model for spectrum sensing stems from the observation that the tails of probability density function of receiver noise are typically heavier that that of Gaussian. The proposed method employs a goodness-of-fit test based on an estimate of the entropy in the observations, compares this decision statistic to a threshold chosen based on a given level on the probability of false-alarm, and decides in the favor of the presence of the incumbent when this statistic exceeds a threshold. Through Monte Carlo simulations, we study the performance of the proposed method and show that it outperforms the matched filter-based detector and the energy detector in practical scenarios.
Sutapa Sarkar, R. Muralishankar, and Sanjeev Gurugopinath
IEEE
Cognitive radio is considered as a promising technology for efficient spectrum utilization. Spectrum sensing algorithm detects primary transmission based on the signal energy at the secondary receiver, well above a prefixed threshold under the binary hypothesis testing setup. In this paper, we investigate the performance of the well-known energy detector (ED), considering the distribution of the exact test statistics. We also consider the Gaussian approximation via the central limit theorem. We evaluate probability of overall error for ED, and derive an optimal threshold under the exact and Gaussian approximated statistics with noise variance uncertainty. Furthermore, we apply a dynamic correction mechanism to the optimal threshold with exact statistics of ED to compute the probability of error with noise variance uncertainty. Additionally, through Monte Carlo simulations, we show that the dynamic correction and threshold optimization significantly reduces the probability of error.
Ajey Saligrama, H.G. Ranjani, R. Muralishankar, and H. N. Shankar
IEEE
In this work, we propose an objective measure to assess the amount of musical noise that results from speech enhancement algorithms. The algorithms can result in nonsmooth suppression of background noise which in turn translates to isolated regions of high energy, referred to as musical noise. We propose to identify such regions by combining time-frequency (TF) bins associated through connectivity along with additional properties of these regions such as area, aspect ratio and total energy. The objective measure proposed is based on density of such regions. The effectiveness of the proposed measure is studied by correlating it with subjective assessment of listeners using enhanced speech of various algorithms.
Shashank Satyanarayana, Narendra K.C., Sanjeev Gurugopinath, R. Muralishankar, and R. Kumaraswamy
IEEE
In this paper, we consider the problem of automatic language identification, and propose the viability of front-end features such as MFCC, and the cepstral features derived with multitapered magnitude (MTMAG) and multitapered modified group delay function (MTMOGDF), using the recently proposed convolutional recurrent neural network (CRNN)-based image classifiers. The CRNN exploits both the spatial and temporal information between the image blocks, and within a given image, for the classification task. The images to these CRNN-based classifiers are constructed based on the considered cepstral-domain features. We conduct a detailed set of experiments on the news database with five International languages, along with the Akashvani news data base with ten Indian languages. The insights from the resultant confusion matrices are inconclusive in the favor of a particular feature. In other words, each of the chosen features perform well for only a set of the chosen languages. However, evaluating the average performance metrics such as equal error rate (EER) and F1 score shows that the MTMOGDF-driven CRNN outperforms those that were designed using MFCC and MTMAG features. In particular, it is seen that the MTMOGDF-driven CRNN achieves nearly 3 dB reduction in EER, with a slight loss of performance in F1 score.
Sanjeev Gurugopinath and R. Muralishankar
Institution of Engineering and Technology (IET)
A new goodness-of-fit test for spectrum sensing in cognitive radios under heavy-tailed noise is proposed, based on the geometric power (also called the zero-order statistics) of the received observations. The noise statistics is assumed to follow a symmetric-alpha-stable distribution, motivated by statistics observed in realistic scenarios. The expressions are provided for the test statistic and the asymptotic detection threshold, in terms of the number of observations under the null hypothesis. Through extensive Monte Carlo simulations, the superior performance of the proposed technique over existing non-linear detection techniques is demonstrated, such as the fractional lower-order statistics, zero-memory non-linear and myriad filtering. In addition, the advantages of the proposed technique on experiment-captured data are demonstrated.
Ajey Saligrama, H. G. Ranjani, H. N. Shankar, and R. Muralishankar
IEEE
We explore the utility of Discriminative Random Fields (DRF) in modulation domain for speech enhancement in stationary noise setting. The ideal channel selection framework is used in modulation domain. We extend the 2-D spectro-temporal DRF framework to model 3-D noisy observations in modulation domain. Each Time-(acoustic)Frequency-Modulation(frequency) bin is classified as speech or noise. Thus, a 3-D binary mask is to be inferred. Iterated Conditional Mode is used to estimate the binary mask; the continuity along acoustic and modulation frequencies is captured using Ising model. By capturing the continuity along modulation frequencies, there is no musical noise present in the enhanced speech. The quality of reconstructed speech in different noise settings (additive white Gaussian noise, pink, car and street noise) is measured in terms of average segmental Signal to Noise Ratio, Perceptual Evaluation of Speech Quality (PESQ) and Mean Opinion Score (MOS). We compare and contrast the performance of this modulation domain DRF with that of spectro-temporal DRF and iterative Wiener Filter.
Debayan Ghosh, R Muralishankar, and Sanjeev Gurugopinath
ISCA
We propose a novel voice activity detection (VAD) scheme employing differential entropy at each frequency bin of power spectral estimates of past and present overlapping speech frames. Here, the power spectral estimate is obtained by employing the Bartlett-Welch method. Later, we add entropies across frequency bins and denote this as the frequency domain long-term differential entropy (FLDE). Longterm averaging enhances VAD performance under low signalto-noise-ratio (SNR). We evaluate the performance of proposed FLDE scheme, considering 12 types of noises and 5 different SNRs which are artificially added to speech samples from the SWITCHBOARD corpus. We present VAD performance of FLDE and compare with existing VAD algorithms, such as ITU-T G.729B, likelihood ratio test, long-term signal variability, and long-term spectral flatness measure based algorithms. Finally, we demonstrate that our FLDE-based VAD performs with best average accuracy and speech hit-rate among the VAD algorithms considered for evaluation.
S. Gurugopinath, R. Muralishankar, and H.N. Shankar
Institution of Engineering and Technology (IET)
The authors investigate the spectral flatness measure (SFM)-based spectrum sensing technique for cognitive radios. This scheme exploits the fact that under Gaussian noise, the noise-only observations have flattened spectrum, i.e. more white, when compared with that of the observations containing the incumbent or primary signal; hence, an increased SFM under the null hypothesis. Under the null hypothesis, the authors derive the asymptotic distribution of the test statistic, and the asymptotically optimal detection threshold, with a constraint on the probability of false-alarm. Furthermore, the authors show that this test is robust to the noise variance uncertainty (NVU) and is related to a test based on the entropy in the observed sequence. Through extensive Monte-Carlo simulations, the authors show that the test based on SFM performs better than the existing energy detector and the blind detector, under realistic signal and fading models, all in the presence of NVU. The authors also highlight the practical utility of this technique based on experimental results.
Sanjeev Gurugopinath, Rangarao Muralishankar, and H.N. Shankar
Institute of Electrical and Electronics Engineers (IEEE)
We propose a novel goodness-of-fit detection scheme for spectrum sensing, based on differential entropy in the received observations. The noise distribution is known to deviate from the Gaussian in many practical communication settings. We, therefore, permit that the noise process follows the generalized Gaussian distribution, which subsumes Gaussian and Laplacian as special cases. We obtain, in closed form, the distribution of the test statistic under the null hypothesis and compute the detection threshold that satisfies a constraint on the probability of false alarm. Furthermore, we derive a lower bound on the probability of detection in a general scenario, using the entropy power inequality. Through Monte Carlo simulations, we show that for a class of practically relevant fading channel and primary signal models, especially in low SNR regime, our detector achieves a higher probability of detection than the energy detector and the order statistics-based detector. We also demonstrate that the adverse effect of noise variance uncertainty is much less with the proposed detector compared with that of the energy detector.
Ajey Saligrama, Ranjani H. G., H. N. Shankarz, and R. Muralishankarx
IEEE
Speech enhancement in stationary noise is addressed using the ideal channel selection framework. In order to estimate the binary mask, we propose to classify each time-frequency (T-F) bin of the noisy signal as speech or noise using Discriminative Random Fields (DRF). The DRF function contains two terms - an enhancement function and a smoothing term. On each T-F bin, we propose to use an enhancement function based on likelihood ratio test for speech presence, while Ising model is used as smoothing function for spectro-temporal continuity in the estimated binary mask. The effect of the smoothing function over successive iterations is found to reduce musical noise as opposed to using only enhancement function. The binary mask is inferred from the noisy signal using Iterated Conditional Modes (ICM) algorithm. Sentences from NOIZEUS corpus are evaluated from 0 dB to 15 dB Signal to Noise Ratio (SNR) in 4 kinds of additive noise settings: additive white Gaussian noise, car noise, street noise and pink noise. The reconstructed speech using the proposed technique is evaluated in terms of average segmental SNR, Perceptual Evaluation of Speech Quality (PESQ) and Mean opinion Score (MOS).