Manohar N. Murthi

All-pole modeling of speech based on the minimum variance distortionless response spectrum

We develop more fully all-pole modeling of speech based on the minimum variance distortionless response (MVDR) spectrum. It is shown that MVDR modeling provides a class of all-pole models that are flexible for tackling a wide variety of speech modeling objectives. In particular the high order MVDR spectrum provides a robust model for all types of speech including voiced speech, unvoiced speech, and mixed spectra. Furthermore, it is simply obtained, and is always superior to the linear prediction (LP) spectrum. With its high quality modeling, the high order MVDR spectrum is suitable for use as a high quality reference spectrum, or for applications like speech recognition. In addition, the MVDR model possesses flexibility for developing low order all-pole models suitable for compression applications. In particular; reduced order MVDR all-pole models are shown to often outperform conventional LP filters in modeling all types of speech spectra. For more accurate modeling of a set of speech spectral samples in the frequency domain, MVDR modeling facilitates the development of superior weighted all-pole filters.

Sparse Linear Prediction and Its Applications to Speech Processing

The aim of this paper is to provide an overview of Sparse Linear Prediction, a set of speech processing tools created by introducing sparsity constraints into the linear prediction framework. These tools have shown to be effective in several issues related to modeling and coding of speech signals. For speech analysis, we provide predictors that are accurate in modeling the speech production process and overcome problems related to traditional linear prediction. In particular, the predictors obtained offer a more effective decoupling of the vocal tract transfer function and its underlying excitation, making it a very efficient method for the analysis of voiced speech. For speech coding, we provide predictors that shape the residual according to the characteristics of the sparse encoding techniques resulting in more straightforward coding strategies. Furthermore, encouraged by the promising application of compressed sensing in signal compression, we investigate its formulation and application to sparse linear predictive coding. The proposed estimators are all solutions to convex optimization problems, which can be solved efficiently and reliably using, e.g., interior-point methods. Extensive experimental results are provided to support the effectiveness of the proposed methods, showing the improvements over traditional linear prediction in both speech analysis and coding.

Hidden Markov model-based packet loss concealment for voice over IP

As voice over IP proliferates, packet loss concealment (PLC) at the receiver has emerged as an important factor in determining voice quality of service. Through the use of heuristic variations of signal and parameter repetition and overlap-add interpolation to handle packet loss, conventional PLC systems largely ignore the dynamics of the statistical evolution of the speech signal, possibly leading to perceptually annoying artifacts. To address this problem, we propose the use of hidden Markov models for PLC. With a hidden Markov model (HMM) tracking the evolution of speech signal parameters, we demonstrate how PLC is performed within a statistical signal processing framework. Moreover, we show how the HMM is used to index a specially designed PLC module for the particular signal context, leading to signal-contingent PLC. Simulation examples, objective tests, and subjective listening tests are provided showing the ability of an HMM-based PLC built with a sinusoidal analysis/synthesis model to provide better loss concealment than a conventional PLC based on the same sinusoidal model for all types of speech signals, including onsets and signal transitions

Retrieving Sparse Patterns Using a Compressed Sensing Framework: Applications to Speech Coding Based on Sparse Linear Prediction

Encouraged by the promising application of compressed sensing in signal compression, we investigate its formulation and application in the context of speech coding based on sparse linear prediction. In particular, a compressed sensing method can be devised to compute a sparse approximation of speech in the residual domain when sparse linear prediction is involved. We compare the method of computing a sparse prediction residual with the optimal technique based on an exhaustive search of the possible nonzero locations and the well known Multi-Pulse Excitation, the first encoding technique to introduce the sparsity concept in speech coding. Experimental results demonstrate the potential of compressed sensing in speech coding techniques, offering high perceptual quality with a very sparse approximated prediction residual.

Regenerative cooperative diversity with path selection and equal power consumption in wireless networks

Recently developed cooperative protocol with distributed path selection provides a simple and practical means of achieving full cooperative diversity in wireless networks. While the best path selection method can significantly improve bit error rate (BER) performance, it may cause unequal power consumption among relay nodes, which may reduce the lifetime of energy-constrained networks. A path selection method under the equal power constraint has been developed for the amplifyand- forward (AF) protocol, but there is no such method for the decode-and-forward (DF) protocol. In this paper, we develop a distributed path selection method with an equal power constraint for the DF protocol. We also analyze the BER performance of our path-selection method. Numerical results demonstrate that the proposed method can guarantee equal power consumption, while achieving full diversity as the best path selection method and providing significant performance gain relative to noncooperative communication.

Regularized Linear Prediction of Speech

All-pole spectral envelope estimates based on linear prediction (LP) for speech signals often exhibit unnaturally sharp peaks, especially for high-pitch speakers. In this paper, regularization is used to penalize rapid changes in the spectral envelope, which improves the spectral envelope estimate. Based on extensive experimental evidence, we conclude that regularized linear prediction outperforms bandwidth-expanded linear prediction. The regularization approach gives lower spectral distortion on average, and fewer outliers, while maintaining a very low computational complexity.

Minimum variance distortionless response (MVDR) modeling of voiced speech

In this paper we propose the MVDR method, which is based upon the minimum variance distortionless response (MVDR) spectrum estimation method, for modeling voiced speech. Developed to overcome some of the shortcomings of linear prediction models, the MVDR method provides better models for medium and high pitch voiced speech. The MVDR model is an all-pole model whose spectrum is easily obtained from a modest non-iterative computation involving the linear prediction coefficients thereby retaining some of the computational attractiveness of LPC methods. With the proper choice of filter order, which is dependent on the number of harmonics, the MVDR spectrum models the formants and spectral powers of voiced speech exactly. An efficient reduced model order MVDR method is developed to further enhance its applicability. An extension of the reduced order MVDR method for recovering the correct amplitudes of the harmonics of voiced speech is also presented.

Enhancing sparsity in linear prediction of speech by iteratively reweighted 1-norm minimization

Linear prediction of speech based on 1-norm minimization has already proved to be an interesting alternative to 2-norm minimization. In particular, choosing the 1-norm as a convex relaxation of the 0-norm, the corresponding linear prediction model offers a sparser residual better suited for coding applications. In this paper, we propose a new speech modeling technique based on reweighted 1-norm minimization. The purpose of the reweighted scheme is to overcome the mismatch between 0-norm minimization and 1-norm minimization while keeping the problem solvable with convex estimation tools. Experimental results prove the effectiveness of the reweighted 1-norm minimization, offering better coding properties compared to 1-norm minimization.

Belief theoretic methods for soft and hard data fusion

In many contexts, one is confronted with the problem of extracting information from large amounts of different types soft data (e.g., text) and hard data (from e.g., physics-based sensing systems). In handling hard data, signal and data processing offers a wealth of methods related to modeling, estimation, tracking, and inference tasks. However, soft data present several challenges that necessitate the development of new data processing methods. For example, with suitable statistical natural language processing (NLP) methods, text can be converted into logic statements that are associated with various forms of associated uncertainty related to the credibility of the statement, the reliability of the text source, and so forth. In combining or fusing soft data with either soft or hard data, one must deploy methods that can suitably preserve and update the uncertainty associated with the data, thereby providing uncertainty bounds related to any inferences regarding semantics. Since standard Bayesian probabilistic approaches have problems with suitably handling uncertain logic statements, there is an emerging need for new methods for processing heterogeneous data. In this paper, we describe a framework for fusing soft and hard data based on the Dempster-Shafer (DS) belief theoretic approach which is well-suited to the task of capturing the types of models and uncertain rules that are more typical of soft data. Since the effectiveness of traditional DS methods has been hampered by high computational requirements, we base the processing framework on our new conditional approach to DS theoretic evidence updating and fusion. We address the issue of laying the foundation for a theoretically justifiable, and computationally efficient framework for fusing soft and hard data taking into account the inherent data uncertainty such as reliability and credibility. Moreover, we present an illustrative example that highlights the potential for the DS conditional approach for fusing heterogeneous data.

Toward Efficient Computation of the Dempster–Shafer Belief Theoretic Conditionals

Dempster-Shafer (DS) belief theory provides a convenient framework for the development of powerful data fusion engines by allowing for a convenient representation of a wide variety of data imperfections. The recent work on the DS theoretic (DST) conditional approach, which is based on the Fagin-Halpern (FH) DST conditionals, appears to demonstrate the suitability of DS theory for incorporating both soft (generated by human-based sensors) and hard (generated by physics-based sources) evidence into the fusion process. However, the computation of the FH conditionals imposes a significant computational burden. One reason for this is the difficulty in identifying the FH conditional core, i.e., the set of propositions receiving nonzero support after conditioning. The conditional core theorem (CCT) in this paper redresses this shortcoming by explicitly identifying the conditional focal elements with no recourse to numerical computations, thereby providing a complete characterization of the conditional core. In addition, we derive explicit results to identify those conditioning propositions that may have generated a given conditional core. This “converse” to the CCT is of significant practical value for studying the sensitivity of the updated knowledge base with respect to the evidence received. Based on the CCT, we also develop an algorithm to efficiently compute the conditional masses (generated by FH conditionals), provide bounds on its computational complexity, and employ extensive simulations to analyze its behavior.