Multi-Pattern Viterbi Algorithm for joint decoding of multiple speech patterns

Joint decoding of multiple speech patterns so as to improve speech recognition performance is important, especially in the presence of noise. In this paper, we propose a Multi-Pattern Viterbi algorithm (MPVA) to jointly decode and recognize multiple speech patterns for automatic speech recognition (ASR). The MPVA is a generalization of the Viterbi Algorithm to jointly decode multiple patterns given a Hidden Markov Model (HMM). Unlike the previously proposed two stage Constrained Multi-Pattern Viterbi Algorithm (CMPVA),the MPVA is a single stage algorithm. MPVA has the advantage that it cart be extended to connected word recognition (CWR) and continuous speech recognition (CSR) problems. MPVA is shown to provide better speech recognition performance than the earlier techniques: using only two repetitions of noisy speech patterns (-5 dB SNR, 10% burst noise), the word error rate using MPVA decreased by 28.5%, when compared to using individual decoding. (C) 2010 Elsevier B.V. All rights reserved.

Rapid localization of damage using a circular sensor array and Lamb wave based triangulation

A damage detection and imaging methodology based on symmetry of neighborhood sensor path and similarity of signal patterns with respect to radial paths in a circular array of sensors has been developed It uses information regarding Limb wave propagation along with a triangulation scheme to rapidly locate and quantify the severity of damage without using all of the sensor data. In a plate like structure, such a scheme can be effectively employed besides full field imaging of wave scattering pattern from the damage, if present in the plate. This new scheme is validated experimentally. Hole and corrosion type damages have been detected and quantified using the proposed scheme successfully. A wavelet based cumulative damage index has been studied which shows monotonic sensitivity against the severity of the damage. which is most desired in a Structural Health Monitoring system. (C) 2010 Elsevier Ltd. All rights reserved.

Filter design using radial basis function neural network and genetic algorithm for improved operational health monitoring

The problem of denoising damage indicator signals for improved operational health monitoring of systems is addressed by applying soft computing methods to design filters. Since measured data in operational settings is contaminated with noise and outliers, pattern recognition algorithms for fault detection and isolation can give false alarms. A direct approach to improving the fault detection and isolation is to remove noise and outliers from time series of measured data or damage indicators before performing fault detection and isolation. Many popular signal-processing approaches do not work well with damage indicator signals, which can contain sudden changes due to abrupt faults and non-Gaussian outliers. Signal-processing algorithms based on radial basis function (RBF) neural network and weighted recursive median (WRM) filters are explored for denoising simulated time series. The RBF neural network filter is developed using a K-means clustering algorithm and is much less computationally expensive to develop than feedforward neural networks trained using backpropagation. The nonlinear multimodal integer-programming problem of selecting optimal integer weights of the WRM filter is solved using genetic algorithm. Numerical results are obtained for helicopter rotor structural damage indicators based on simulated frequencies. Test signals consider low order polynomial growth of damage indicators with time to simulate gradual or incipient faults and step changes in the signal to simulate abrupt faults. Noise and outliers are added to the test signals. The WRM and RBF filters result in a noise reduction of 54 - 71 and 59 - 73% for the test signals considered in this study, respectively. Their performance is much better than the moving average FIR filter, which causes significant feature distortion and has poor outlier removal capabilities and shows the potential of soft computing methods for specific signal-processing applications. (C) 2005 Elsevier B. V. All rights reserved.

Epoch extraction from linear prediction residual for identification of closed glottis interval

In voiced speech analysis epochal information is useful in accurate estimation of pitch periods and the frequency response of the vocal tract system. Ideally, linear prediction (LP) residual should give impulses at epochs. However, there are often ambiguities in the direct use of LP residual since samples of either polarity occur around epochs. Further, since the digital inverse filter does not compensate the phase response of the vocal tract system exactly, there is an uncertainty in the estimated epoch position. In this paper we present an interpretation of LP residual by considering the effect of the following factors: 1) the shape of glottal pulses, 2) inaccurate estimation of formants and bandwidths, 3) phase angles of formants at the instants of excitation, and 4) zeros in the vocal tract system. A method for the unambiguous identification of epochs from LP residual is then presented. The accuracy of the method is tested by comparing the results with the epochs obtained from the estimated glottal pulse shapes for several vowel segments. The method is used to identify the closed glottis interval for the estimation of the true frequency response of the vocal tract system.

Signal characterization using fractal dimension

Fractal Dimensions (FD) are one of the popular measures used for characterizing signals. They have been used as complexity measures of signals in various fields including speech and biomedical applications. However, proper interpretation of such analyses has not been thoroughly addressed. In this paper, we study the effect of various signal properties on FD and interpret results in terms of classical signal processing concepts such as amplitude, frequency, number of harmonics, noise power and signal bandwidth. We have used Higuchi's method for estimating FDs. This study may help in gaining a better understanding of the FD complexity measure itself, and for interpreting changing structural complexity of signals in terms of FD. Our results indicate that FD is a useful measure in quantifying structural changes in signal properties.

A Sub-Nyquist Sampling Method For Computing The Level-crossing-times Of An Analog Signal: Theory And Applications

We address the problem of computing the level-crossings of an analog signal from samples measured on a uniform grid. Such a problem is important, for example, in multilevel analog-to-digital (A/D) converters. The first operation in such sampling modalities is a comparator, which gives rise to a bilevel waveform. Since bilevel signals are not bandlimited, measuring the level-crossing times exactly becomes impractical within the conventional framework of Shannon sampling. In this paper, we propose a novel sub-Nyquist sampling technique for making measurements on a uniform grid and thereby for exactly computing the level-crossing times from those samples. The computational complexity of the technique is low and comprises simple arithmetic operations. We also present a finite-rate-of-innovation sampling perspective of the proposed approach and also show how exponential splines fit in naturally into the proposed sampling framework. We also discuss some concrete practical applications of the sampling technique.

Parts based representation for pedestrian using NMF with robustness to partial occlusion

Computer Vision has seen a resurgence in the parts-based representation for objects over the past few years. The parts are usually annotated beforehand for training. We present an annotation free parts-based representation for the pedestrian using Non-Negative Matrix Factorization (NMF). We show that NMF is able to capture the wide range of pose and clothing of the pedestrians. We use a modified form of NMF i.e. NMF with sparsity constraints on the factored matrices. We also make use of Riemannian distance metric for similarity measurements in NMF space as the basis vectors generated by NMF aren't orthogonal. We show that for 1% drop in accuracy as compared to the Histogram of Oriented Gradients (HOG) representation we can achieve robustness to partial occlusion.

A Study of Long-range Correlations in Schizophrenia EEG using Detrended Fluctuation Analysis

In this paper, we have studied electroencephalogram (EEG) activity of schizophrenia patients, in resting eyes closed condition, with detrended fluctuation analysis (DFA). The DFA gives information about scaling and long-range correlations in time series. We computed DFA exponents from 30 scalp locations of 18 male neuroleptic-naIve, recent-onset schizophrenia (NRS) subjects and 15 healthy male control subjects. Our results have shown two scaling regions in all the scalp locations in all the subjects, with different slopes, corresponding to two scaling exponents. No significant differences between the groups were found with first scaling exponent (short-range). However, the second scaling exponent (long-range) were significantly lower in control subjects at all scalp locations (p<0.05, Kruskal-Wallis test). These findings suggest that the long-range scaling behavior of EEG is sensitive to schizophrenia, and this may provide an additional insight into the brain dysfunction in schizophrenia.

MOSAICS: Multiplexed Optimal Signal Acquisition Involving Compressed Sensing

It is possible to sample signals at sub-Nyquist rate and still be able to reconstruct them with reasonable accuracy provided they exhibit local Fourier sparsity. Underdetermined systems of equations, which arise out of undersampling, have been solved to yield sparse solutions using compressed sensing algorithms. In this paper, we propose a framework for real time sampling of multiple analog channels with a single A/D converter achieving higher effective sampling rate. Signal reconstruction from noisy measurements on two different synthetic signals has been presented. A scheme of implementing the algorithm in hardware has also been suggested.

Image analysis using multifractals

The efficacy of the multifractal spectrum as a tool for characterizing images has been studied. This spectrum has been computed for digitized images of the nucleus of human cervical cancer cells and it was observed that the entire spectrum is almost fully reproduced for a normal cell while only the right half (q<0) of the spectrum is reproduced for a cancerous cell. Cells in stages in between the two extremes show a shortening of the left half of the spectrum proportional to their condition. The extent of this shortening has been found to be sufficient to permit a classification between three classes of cells at varying distances from a basal cancerous layer-the superficial cells, the intermediate cells and the parabasal cells. This technique may be used for automatic screening of the population while also indicating the stage of malignancy

Some Algorithms for Eigensubspace Estimation

An important tool in signal processing is the use of eigenvalue and singular value decompositions for extracting information from time-series/sensor array data. These tools are used in the so-called subspace methods that underlie solutions to the harmonic retrieval problem in time series and the directions-of-arrival (DOA) estimation problem in array processing. The subspace methods require the knowledge of eigenvectors of the underlying covariance matrix to estimate the parameters of interest. Eigenstructure estimation in signal processing has two important classes: (i) estimating the eigenstructure of the given covariance matrix and (ii) updating the eigenstructure estimates given the current estimate and new data. In this paper, we survey some algorithms for both these classes useful for harmonic retrieval and DOA estimation problems. We begin by surveying key results in the literature and then describe, in some detail, energy function minimization approaches that underlie a class of feedback neural networks. Our approaches estimate some or all of the eigenvectors corresponding to the repeated minimum eigenvalue and also multiple orthogonal eigenvectors corresponding to the ordered eigenvalues of the covariance matrix. Our presentation includes some supporting analysis and simulation results. We may point out here that eigensubspace estimation is a vast area and all aspects of this cannot be fully covered in a single paper. (C) 1995 Academic Press, Inc.

Exact complex-wave reconstruction in digital holography

We address the problem of exact complex-wave reconstruction in digital holography. We show that, by confining the object-wave modulation to one quadrant of the frequency domain, and by maintaining a reference-wave intensity higher than that of the object, one can achieve exact complex-wave reconstruction in the absence of noise. A feature of the proposed technique is that the zero-order artifact, which is commonly encountered in hologram reconstruction, can be completely suppressed in the absence of noise. The technique is noniterative and nonlinear. We also establish a connection between the reconstruction technique and homomorphic signal processing, which enables an interpretation of the technique from the perspective of deconvolution. Another key contribution of this paper is a direct link between the reconstruction technique and the two-dimensional Hilbert transform formalism proposed by Hahn. We show that this connection leads to explicit Hilbert transform relations between the magnitude and phase of the complex wave encoded in the hologram. We also provide results on simulated as well as experimental data to validate the accuracy of the reconstruction technique. (C) 2011 Optical Society of America

Performance analysis of a modified spatial smoothing technique for direction estimation

The statistical performance analysis of ESPRIT, root-MUSIC, minimum-norm methods for direction estimation, due to finite data perturbations, using the modified spatially smoothed covariance matrix, is developed. Expressions for the mean-squared error in the direction estimates are derived based on a common framework. Based on the analysis, the use of the modified smoothed covariance matrix improves the performance of the methods when the sources are fully correlated. Also, the performance is better even when the number of subarrays is large unlike in the case of the conventionally smoothed covariance matrix. However, the performance for uncorrelated sources deteriorates due to an artificial correlation introduced by the modified smoothing. The theoretical expressions are validated using extensive simulations. (C) 1999 Elsevier Science B.V. All rights reserved.

Strip based coding for large images using wavelets

Thanks to advances in sensor technology, today we have many applications (space-borne imaging, medical imaging, etc.) where images of large sizes are generated. Straightforward application of wavelet techniques for above images involves certain difficulties. Embedded coders such as EZW and SPIHT require that the wavelet transform of the full image be buffered for coding. Since the transform coefficients also require storing in high precision, buffering requirements for large images become prohibitively high. In this paper, we first devise a technique for embedded coding of large images using zero trees with reduced memory requirements. A 'strip buffer' capable of holding few lines of wavelet coefficients from all the subbands belonging to the same spatial location is employed. A pipeline architecure for a line implementation of above technique is then proposed. Further, an efficient algorithm to extract an encoded bitstream corresponding to a region of interest in the image has also been developed. Finally, the paper describes a strip based non-embedded coding which uses a single pass algorithm. This is to handle high-input data rates. (C) 2002 Elsevier Science B.V. All rights reserved.

Speech segmentation using extrema based signal track length measure

We introduce a novel temporal feature of a signal, namely extrema-based signal track length (ESTL) for the problem of speech segmentation. We show that ESTL measure is sensitive to both amplitude and frequency of the signal. The short-time ESTL (ST_ESTL) shows a promising way to capture the significant segments of speech signal, where the segments correspond to acoustic units of speech having distinct temporal waveforms. We compare ESTL based segmentation with ML and STM methods and find that it is as good as spectral feature based segmentation, but with lesser computational complexity.