Decision tree design and applications in speech processing

The design and operation of the minimum cost classifier, where the total cost is the sum of the measurement cost and the classification cost, is computationally complex. Noting the difficulties associated with this approach, decision tree design directly from a set of labelled samples is proposed in this paper. The feature space is first partitioned to transform the problem to one of discrete features. The resulting problem is solved by a dynamic programming algorithm over an explicitly ordered state space of all outcomes of all feature subsets. The solution procedure is very general and is applicable to any minimum cost pattern classification problem in which each feature has a finite number of outcomes. These techniques are applied to (i) voiced, unvoiced, and silence classification of speech, and (ii) spoken vowel recognition. The resulting decision trees are operationally very efficient and yield attractive classification accuracies.

Fast Likelihood Computation in Speech Recognition using Matrices

Acoustic modeling using mixtures of multivariate Gaussians is the prevalent approach for many speech processing problems. Computing likelihoods against a large set of Gaussians is required as a part of many speech processing systems and it is the computationally dominant phase for Large Vocabulary Continuous Speech Recognition (LVCSR) systems. We express the likelihood computation as a multiplication of matrices representing augmented feature vectors and Gaussian parameters. The computational gain of this approach over traditional methods is by exploiting the structure of these matrices and efficient implementation of their multiplication. In particular, we explore direct low-rank approximation of the Gaussian parameter matrix and indirect derivation of low-rank factors of the Gaussian parameter matrix by optimum approximation of the likelihood matrix. We show that both the methods lead to similar speedups but the latter leads to far lesser impact on the recognition accuracy. Experiments on 1,138 work vocabulary RM1 task and 6,224 word vocabulary TIMIT task using Sphinx 3.7 system show that, for a typical case the matrix multiplication based approach leads to overall speedup of 46 % on RM1 task and 115 % for TIMIT task. Our low-rank approximation methods provide a way for trading off recognition accuracy for a further increase in computational performance extending overall speedups up to 61 % for RM1 and 119 % for TIMIT for an increase of word error rate (WER) from 3.2 to 3.5 % for RM1 and for no increase in WER for TIMIT. We also express pairwise Euclidean distance computation phase in Dynamic Time Warping (DTW) in terms of matrix multiplication leading to saving of approximately of computational operations. In our experiments using efficient implementation of matrix multiplication, this leads to a speedup of 5.6 in computing the pairwise Euclidean distances and overall speedup up to 3.25 for DTW.

Detection of the closure-burst transitions of stops and affricates in continuous speech using the plosion index

Automatic and accurate detection of the closure-burst transition events of stops and affricates serves many applications in speech processing. A temporal measure named the plosion index is proposed to detect such events, which are characterized by an abrupt increase in energy. Using the maxima of the pitch-synchronous normalized cross correlation as an additional temporal feature, a rule-based algorithm is designed that aims at selecting only those events associated with the closure-burst transitions of stops and affricates. The performance of the algorithm, characterized by receiver operating characteristic curves and temporal accuracy, is evaluated using the labeled closure-burst transitions of stops and affricates of the entire TIMIT test and training databases. The robustness of the algorithm is studied with respect to global white and babble noise as well as local noise using the TIMIT test set and on telephone quality speech using the NTIMIT test set. For these experiments, the proposed algorithm, which does not require explicit statistical training and is based on two one-dimensional temporal measures, gives a performance comparable to or better than the state-of-the-art methods. In addition, to test the scalability, the algorithm is applied on the Buckeye conversational speech corpus and databases of two Indian languages. (C) 2014 Acoustical Society of America.

A fast algorithm for speech polarity detection using long-term linear prediction

Speech polarity detection is a crucial first step in many speech processing techniques. In this paper, an algorithm is proposed that improvises the existing technique using the skewness of the voice source (VS) signal. Here, the integrated linear prediction residual (ILPR) is used as the VS estimate, which is obtained using linear prediction on long-term frames of the low-pass filtered speech signal. This excludes the unvoiced regions from analysis and also reduces the computation. Further, a modified skewness measure is proposed for decision, which also considers the magnitude of the skewness of the ILPR along with its sign. With the detection error rate (DER) as the performance metric, the algorithm is tested on 8 large databases and its performance (DER=0.20%) is found to be comparable to that of the best technique (DER=0.06%) on both clean and noisy speech. Further, the proposed method is found to be ten times faster than the best technique.

Bayesian and Decision Tree Approaches for Pattern Recognition Including Feature Measurement Costs

The minimum cost classifier when general cost functionsare associated with the tasks of feature measurement and classification is formulated as a decision graph which does not reject class labels at intermediate stages. Noting its complexities, a heuristic procedure to simplify this scheme to a binary decision tree is presented. The optimizationof the binary tree in this context is carried out using ynamicprogramming. This technique is applied to the voiced-unvoiced-silence classification in speech processing.

Analysis of ECG by multipulse excited linear prediction model

He propose a new time domain method for efficient representation of the KCG and delineation of its component waves. The method is based on the multipulse Linear prediction (LP) coding which is being widely used in speech processing. The excitation to the LP synthesis filter consists of a few pulses defined by their locations and amplitudes. Based on the amplitudes and their distribution, the pulses are suitably combined to delineate the component waves. Beat to beat correlation in the ECG signal is used in QRS periodicity prediction. The method entails a data compression of 1 in 6. The method reconstructs the signal with an NMSE of less than 5%.

Fast acoustic likelihood computation using low-rank matrix approximation

Acoustic modeling using mixtures of multivariate Gaussians is the prevalent approach for many speech processing problems. Computing likelihoods against a large set of Gaussians is required as a part of many speech processing systems and it is the computationally dominant phase for LVCSR systems. We express the likelihood computation as a multiplication of matrices representing augmented feature vectors and Gaussian parameters. The computational gain of this approach over traditional methods is by exploiting the structure of these matrices and efficient implementation of their multiplication.In particular, we explore direct low-rank approximation of the Gaussian parameter matrix and indirect derivation of low-rank factors of the Gaussian parameter matrix by optimum approximation of the likelihood matrix. We show that both the methods lead to similar speedups but the latter leads to far lesser impact on the recognition accuracy. Experiments on a 1138 word vocabulary RM1 task using Sphinx 3.7 system show that, for a typical case the matrix multiplication approach leads to overall speedup of 46%. Both the low-rank approximation methods increase the speedup to around 60%, with the former method increasing the word error rate (WER) from 3.2% to 6.6%, while the latter increases the WER from 3.2% to 3.5%.

Spectral Zero-Crossings: Localization Properties and Applications

We present an analysis of the rate of sign changes in the discrete Fourier spectrum of a sequence. The sign changes of either the real or imaginary parts of the spectrum are considered, and the rate of sign changes is termed as the spectral zero-crossing rate (SZCR). We show that SZCR carries information pertaining to the locations of transients within the temporal observation window. We show duality with temporal zero-crossing rate analysis by expressing the spectrum of a signal as a sum of sinusoids with random phases. This extension leads to spectral-domain iterative filtering approaches to stabilize the spectral zero-crossing rate and to improve upon the location estimates. The localization properties are compared with group-delay-based localization metrics in a stylized signal setting well-known in speech processing literature. We show applications to epoch estimation in voiced speech signals using the SZCR on the integrated linear prediction residue. The performance of the SZCR-based epoch localization technique is competitive with the state-of-the-art epoch estimation techniques that are based on average pitch period.

Efficient post-processing techniques for speech enhancement

We address the problem of speech enhancement in real-world noisy scenarios. We propose to solve the problem in two stages, the first comprising a generalized spectral subtraction technique, followed by a sequence of perceptually-motivated post-processing algorithms. The role of the post-processing algorithms is to compensate for the effects of noise as well as to suppress any artifacts created by the first-stage processing. The key post-processing mechanisms are aimed at suppressing musical noise and to enhance the formant structure of voiced speech as well as to denoise the linear-prediction residual. The parameter values in the techniques are fixed optimally by experimentally evaluating the enhancement performance as a function of the parameters. We used the Carnegie-Mellon university Arctic database for our experiments. We considered three real-world noise types: fan noise, car noise, and motorbike noise. The enhancement performance was evaluated by conducting listening experiments on 12 subjects. The listeners reported a clear improvement (MOS improvement of 0.5 on an average) over the noisy signal in the perceived quality (increase in the mean-opinion score (MOS)) for positive signal-to-noise-ratios (SNRs). For negative SNRs, however, the improvement was found to be marginal.

A post processing technique for improved harmonic/percussion separation in polyphonic speech

In this paper we propose a postprocessing technique for a spectrogram diffusion based harmonic/percussion decom- position algorithm. The proposed technique removes har- monic instrument leakages in the percussion enhanced out- puts of the baseline algorithm. The technique uses median filtering and an adaptive detection of percussive segments in subbands followed by piecewise signal reconstruction using envelope properties to ensure that percussion is enhanced while harmonic leakages are suppressed. A new binary mask is created for the percussion signal which upon applying on the original signal improves harmonic versus percussion separation. We compare our algorithm with two recent techniques and show that on a database of polyphonic Indian music, the postprocessing algorithm improves the harmonic versus percussion decomposition significantly.

Compressive Sensing For Sparsely Excited Speech Signals

Compressive sensing (CS) has been proposed for signals with sparsity in a linear transform domain. We explore a signal dependent unknown linear transform, namely the impulse response matrix operating on a sparse excitation, as in the linear model of speech production, for recovering compressive sensed speech. Since the linear transform is signal dependent and unknown, unlike the standard CS formulation, a codebook of transfer functions is proposed in a matching pursuit (MP) framework for CS recovery. It is found that MP is efficient and effective to recover CS encoded speech as well as jointly estimate the linear model. Moderate number of CS measurements and low order sparsity estimate will result in MP converge to the same linear transform as direct VQ of the LP vector derived from the original signal. There is also high positive correlation between signal domain approximation and CS measurement domain approximation for a large variety of speech spectra.

GMM based Bayesian approach to speech enhancement in signal transform domain

Considering a general linear model of signal degradation, by modeling the probability density function (PDF) of the clean signal using a Gaussian mixture model (GMM) and additive noise by a Gaussian PDF, we derive the minimum mean square error (MMSE) estimator. The derived MMSE estimator is non-linear and the linear MMSE estimator is shown to be a special case. For speech signal corrupted by independent additive noise, by modeling the joint PDF of time-domain speech samples of a speech frame using a GMM, we propose a speech enhancement method based on the derived MMSE estimator. We also show that the same estimator can be used for transform-domain speech enhancement.

A speech-music discriminator using HILN model based features

We propose a simple speech music discriminator that uses features based on HILN(Harmonics, Individual Lines and Noise) model. We have been able to test the strength of the feature set on a standard database of 66 files and get an accuracy of around 97%. We also have tested on sung queries and polyphonic music and have got very good results. The current algorithm is being used to discriminate between sung queries and played (using an instrument like flute) queries for a Query by Humming(QBH) system currently under development in the lab.

Dynamic programming based optimum non-uniform samples for speech reconstruction and coding

Non-uniform sampling of a signal is formulated as an optimization problem which minimizes the reconstruction signal error. Dynamic programming (DP) has been used to solve this problem efficiently for a finite duration signal. Further, the optimum samples are quantized to realize a speech coder. The quantizer and the DP based optimum search for non-uniform samples (DP-NUS) can be combined in a closed-loop manner, which provides distinct advantage over the open-loop formulation. The DP-NUS formulation provides a useful control over the trade-off between bitrate and performance (reconstruction error). It is shown that 5-10 dB SNR improvement is possible using DP-NUS compared to extrema sampling approach. In addition, the close-loop DP-NUS gives a 4-5 dB improvement in reconstruction error.

Comparative study of filter-bank mean-energy distance for automated segmentation of speech signals

This paper describes a method of automated segmentation of speech assuming the signal is continuously time varying rather than the traditional short time stationary model. It has been shown that this representation gives comparable if not marginally better results than the other techniques for automated segmentation. A formulation of the 'Bach' (music semitonal) frequency scale filter-bank is proposed. A comparative study has been made of the performances using Mel, Bark and Bach scale filter banks considering this model. The preliminary results show up to 80 % matches within 20 ms of the manually segmented data, without any information of the content of the text and without any language dependence. 'Bach' filters are seen to marginally outperform the other filters.