We present an analysis on multi-user pairwise error probability (PEP) performance of non-orthogonal multiple access (NOMA)-aided power line communication (PLC) systems, under heavy-tailed noise. In contrast to the typical assumption in the literature, we consider the successive interference cancellation procedure in NOMA to be imperfect. We model the heavy-tailed noise as generalized Gaussian (GG) distribution with a heavy-tail parameter, and discuss the impact of this parameter on the PEP performance. We validate our analysis using Monte Carlo simulations, and disucss the impact of the lognormal fading parameters in PEP. Further, we numerically show that the diversity order on PEP is equal to the user order for larger values of the heavy-tail parameter, which is similar to that observed in an additive white Gaussian noise scenario.

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.

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 sub-sampled 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 micro-controller. TED can be easily adopted on satellites to decode signals for space Internet of Things applications.

Cognitive Radios (CRs) address spectrum scarcity and under-utilization of the spectrum. Allocating spectrum to CR nodes in a network is neither easy nor straight-forward since spectrum availability is space and time-varying. The Link layer always sets up communication between nodes in a CR ad hoc network by allocating the available spectrum suitably. In a time-slotted model, allocation of channels amongst CR nodes is a NP complete problem. Moreover, since the availability of spectrum is time varying, all the nodes in the network may not be allocated sufficient resources. Thus there is always a possibility that a node may starve compared to its counterparts. In this article, we address the issue of fairness in a CR ad hoc network and propose novel distributed heuristics to allocate spectrum fairly. In literature, max-min, max-sum and max-proportional fairness based schemes are proposed to provide fairness in these CR ad hoc networks. We compare our heuristic results with these schemes, and discuss why our method is significantly better.

In a distributed multiple sensor multiple antenna setup, transmission over a fading multiple access channel results in incoherent fusion of sensor observation. With an overall sensor power constraint, uniform power allocation (PA) over the sensors is optimal when the channel between the sensors and the fusion center (FC) is Additive White Gaussian Noise (AWGN). However, with varying observation signal-to-noise ratios (SNR) at the sensors, uniform PA is not optimal even with AWGN channel. In this paper, we consider imperfect communication between the sensors and the FC and correlated noise at the sensors. We propose a joint power optimization scheme that maximizes the probability of detection (PD) at the FC, subject to an overall power constraint. Also, we derive a closed form expression for the optimization problem when the channel is AWGN. We verify the performance of our algorithm by conducting numerical experiments. We show that the maximum power is allocated to a sensor which is operating with the best SNR and channel condition. We further show that with optimal PA, we never unduly drain our on-board sensor energy and hence extend the battery life without compromising on the PD. Finally, we show that with optimal PA, the total system power requirement is brought down to nearly 10% of that with uniform PA, even when achieving the same PD.

Realtime voice communication over the Internet has rapidly gained popularity. It is indeed essential to reduce the total bandwidth consumption to efficiently use the available bandwidth for the subscribers having low speed connectivity and even otherwise. In this paper we introduce a novel technique to identify the voice and silent regions of a speech stream that is very much suitable for VoIP calls. We use an entropy measure, which is based on the spacings of order statistics of speech frames to differentiate the silence zones from the speech zones. We developed an algorithm that uses an adaptive thresholding to minimize the misdetection. The performance of our approach is compared with the built-in VAD of AMR codec. Our approach yields comparatively better saving in bandwidth yet maintaining a good quality of the speech streams. Further, the proposed approach has improved voice detection compared to the AMR schemes under noisy conditions. The ideas presented in this paper has been identified novel during the WIPO international patent search.

VoIP packets, when transported over the Internet, experience loss and variable delay. The effect of the network not only depends on the background flows but also on the parameters of VoIP packets itself, such as VoIP packet size and the packet generation intervals. While higher sized packets experience more losses, they experience less delay jitter and handling them is thus easy at the playout buffer. To investigate the effect of various network conditions on VoIP session holistically, we present a complete end to end study considering various states of the underlying network. We present as a case study of G.711 coded packets generated at 20 and 40 ms intervals for comparison. While packets carrying 20 ms data are better when the network is loaded, 40 ms packetization is favored when the network is not saturated. This affects the jitter and loss thus affecting the quality. We explain this trade-off using mean opinion scores.

In our earlier work [1, 2], we employed minimum variance distortionless response (MVDR) and MVDR Bauer respectively, as spectral estimation techniques in place of modified-linear prediction in Discrete cosine transform (DCT) based pitch modification [3]. As a general extension, we introduce psychoacoustic characteristics to [1, 2] resulting in Perceptual-MVDR (PMVDR) and PMVDR-Bauer algorithms utilized here for spectral estimation. Further, we employ Bauer method of spectral factorization in our later algorithm since it results in causal inverse filter. These are used to obtain residual signal from pitch synchronous speech frames. The residual signal is resampled using DCT/IDCT depending on the target pitch scale factor. Finally, forward filters realized from the above factorization are used to get pitch modified speech. The modified speech is evaluated subjectively by 10 listeners and mean opinion scores (MOS) are evaluated for pitch factors from 0.5 to 2. Modified bark spectral distortion (MBSD) measure is also employed to evaluate objective performance. We found that the proposed approach has been rated with higher MOS and has achieved lower MBSD than the time domain pitch synchronous overlap [4], modified-LP method [3] and MVDR based methods [1, 2]. Further, we modified the pitch contours of 20 affirmative sentences to sound like interrogative sentences, using the current as well as our earlier algorithms and compared their performance.

In this letter, we derive the theoretical complex cepstrum (TCC) of the discrete cosine transform (DCT) and warped DCT (WDCT) filters. Using these derivations, we intend to develop an analytic model of the warped discrete cosine transform cepstrum (WDCTC), which was recently introduced as a speech processing feature. In our derivation, we start with the filter bank structure for the DCT, where each basis is represented by a finite impulse response (FIR) filter. The WDCT filter bank is obtained by substituting z-1in the DCT filter bank with a first-order all-pass filter. Using the filter bank structures, we first derive the transfer functions for the DCT and WDCT, and subsequently, the TCC for each filter is computed. We analyze the DCT and WDCT filter transfer functions and the TCC by illustrating the corresponding pole-zero maps and cepstral sequences. Moreover, we also use the derived TCC expressions to compute the cepstral sequence for a synthetic vowel /aa/ where the observations on the theoretical cepstrum corroborate well with our practical findings

Voice enabled applications over the Internet are rapidly gaining popularity. Reducing the total bandwidth requirement can make a non-trivial difference for the subscribers having low speed connectivity. Voice activity detection algorithms for VoIP applications can save bandwidth by filtering the frames that do not contain speech. In this paper we introduce a novel technique to identify the voice and silent regions of a speech stream very much suitable for VoIP calls. We use an information theoretic measure, called spectral entropy, for differentiating the silence from the speech zones. Specifically we developed a heuristic approach that uses an adaptive threshold to minimize the miss detection in the presence of noise. The performance of our approach is compared with the relatively new 3GPP TS 26.194 (AMR-WB) standard, along with the listeners' intelligibility rating. Our algorithm yields comparatively better saving in bandwidth, yet maintaining good quality of the speech streams.

The subspaces representing two different vowel classes may have a large common subspace due to speaker variability, noise and coarticulation. We use common principal component (CPC) and its extension i.e., partial-CPC (pCPC) to obtain a specific subspace for each vowel which is insensitive to variations. We perform CPC analysis on the covariance matrices of the vowels. An input vector from an unknown vowel is classified based on the maximum length of its projection on the specific subspaces. We have chosen 18D mel-frequency cepstral coefficients as a feature in our recognition task. The specific subspace is treated as a transformation matrix which enhances the vowel-specific information in the feature vector and, in turn, increases SNR. Experiments were performed on vowels extracted from a multiple speaker set taken from different dialect regions. The results are encouraging in the context of a speaker-independent framework. Visual analysis of the vowel basis spectra provides useful and interesting information by highlighting the importance of different frequency regions.

Abstract In this paper, we continue our investigation of the warpeddiscrete cosine transform cepstrum (WDCTC), which wasearlier introduced as a new speech processing feature [1].Here, we study the statistical properties of the WDCTC andcompare them with the mel-frequency cepstral coefficients(MFCC). We report some interesting properties of the WD-CTC when compared to the MFCC: its statistical distrib-ution is more Gaussian-like with lower variance, it obtainsbettervowel clusterseparability,it formstightervowelclus-ters and generates better codebooks. Further, we employthe WDCTC and MFCC features in a 5-vowel recognitiontask using Vector Quantization (VQ) and 1-Nearest Neigh-bour(1-NN)as classifiers. Inourexperiments,the WDCTCconsistently outperforms the MFCC. 1. Introduction We recently introduced the warped discrete cosine trans-form cepstrum (WDCTC) as a new speech processing fea-ture and demonstrated its better performance than the mel-frequency cepstral coefficients (MFCC) in a vowel recog-nition and speaker-identification task [1]. The WDCTC hasshown good promise as a speech processing feature and weare encouraged to further investigate the WDCTC featureand its statistical properties.Alargevolumeoftrainingdataisrequiredto buildspea-ker-independentspeechrecognitionsystems. Onetechniqueof reducing the data size is clustering the data and choos-ing a reasonable number of representative feature vectorsto form codebooks [2]. Hence, codebook techniques arevery relevant and practical to speech recognition systems.We form WDCTC and MFCC codebooks using a k-meansclustering algorithm and compare the codebook statisticsfor clean and noisy vowels using the coefficient of varianceand overlap ratio (defined later). Our experiment demon-strates that the WDCTC codebooks represent the underly-ing vowel data better than MFCC.In order to compare the classification capability of thefeatures, the WDCTC and MFCC are employed in a 5-vowel recognition task. Vector quantization (VQ) and 1-nearestneighbor(1-NN,[2])are usedas classifiersandtheirrecognition performance is reported. We also investigatethe clean and noisy vowel clusters formed by WDCTC andMFCC features and present the average separability of thevowel classes.

We propose a new technique for modifying the time-scale of speech using Independent Subspace Analysis (ISA). To carry out ISA, the single channel mixture signal is converted to a time-frequency rep-resentation such as spectrogram. Here, the spectrogram is gener-ated by taking Hartley or Wavelet transform on overlapped frames of speech. We do dimensionality reduction of the autocorrelated original spectrogram using singular value decomposition. Then, we use Independent component analysis to get unmixing matrix using JadeICA algorithm [5]. It is then assumed that the over-all spectrogram results from the superposition of a number of un-known statistically independent spectrograms. By using unmix-ing matrix, independent sources such as temporal amplitude en-velopes and frequency weights can be extracted from the spec-trogram. Time-scaling of speech is carried out by resampling the independent temporal amplitude envelopes. We then obtain time-scaled independent spectrograms after multiplying the independent frequency weights with time-scaled temporal amplitude envelopes. Summing all these independent spectrograms and taking inverse Hartely or wavelet transform of the sum spectrogram to recon-struct and overlap-add the reconstructed time-domain signal to get the time-scaled speech. The quality of the time-scaled speech has been analyzed using Modified Bark Spectral Distortion(MBSD) [6]. From the MBSD score, one can infer that the time-scaled sig-nal is less distorted. 1.

We propose two algorithms. which effectively use Discrete Cosine Transform (DCT) to obtain complex cepstrum, which we call as the Discrete Cosine Transformed Cepstrum (DCTC). The first algorithm uses the relation between Discrete Fourier Transform (DFI) and DCT II. Computing the complex cepstrum using Fourier Transform needs unwrapped phase. The calculation of the unwrapped phase is sometimes difficult due to the presence of multiple zeros and poles near or on the unit circle. Since DCT is a real function, its phase can only be 0 or π and phase unwrapping is done by representing negative by exp( ∼jπ) and positive by 1. The motivation for the second algorithm is to obtain DCTC by representing DCT sequence itself by magnitude and phase components. Phase unwrapping is performed in the same way as in the case of first algorithm. We tested DCTC by simulating the impulse response of a system that has multiple poles and zeros near or on the unit circle and we have shown that DCTC provides a closer match to the theoretical complex cepstrum than the DFT based complex cepstrum. DCTC has been used to obtain the pitch contour of syllables, words and sentences. It is shown that the first few coefficients' spectral envelope matches with the envelope of the signal spectrum under consideration. We are also exploring the applicability of DCTC to speaker recognition.