Kuldip Kumar Paliwal

Bidirectional recurrent neural networks

In the first part of this paper, a regular recurrent neural network (RNN) is extended to a bidirectional recurrent neural network (BRNN). The BRNN can be trained without the limitation of using input information just up to a preset future frame. This is accomplished by training it simultaneously in positive and negative time direction. Structure and training procedure of the proposed network are explained. In regression and classification experiments on artificial data, the proposed structure gives better results than other approaches. For real data, classification experiments for phonemes from the TIMIT database show the same tendency. In the second part of this paper, it is shown how the proposed bidirectional structure can be easily modified to allow efficient estimation of the conditional posterior probability of complete symbol sequences without making any explicit assumption about the shape of the distribution. For this part, experiments on real data are reported.

Efficient vector quantization of LPC parameters at 24 bits/frame

For low bit rate speech coding applications, it is important to quantize the LPC parameters accurately using as few bits as possible. Though vector quantizers are more efficient than scalar quantizers, their use for accurate quantization of linear predictive coding (LPC) information (using 24-26 bits/frame) is impeded by their prohibitively high complexity. A split vector quantization approach is used here to overcome the complexity problem. An LPC vector consisting of 10 line spectral frequencies (LSFs) is divided into two parts, and each part is quantized separately using vector quantization. Using the localized spectral sensitivity property of the LSF parameters, a weighted LSF distance measure is proposed. With this distance measure, it is shown that the split vector quantizer can quantize LPC information in 24 bits/frame with an average spectral distortion of 1 dB and less than 2% of the frames having spectral distortion greater than 2 dB. The effect of channel errors on the performance of this quantizer is also investigated and results are reported. >

A speech enhancement method based on Kalman filtering

In this paper, the problem of speech enhancement when only corrupted speech signal is available for processing is considered. For this, the Kalman filtering method is studied and compared with the Wiener filtering method. Its performance is found to be significantly better than the Wiener filtering method. A delayed-Kalman filtering method is also proposed which improves the speech enhancement performance of Kalman filter further.

Feature extraction and dimensionality reduction algorithms and their applications in vowel recognition

Feature extraction is an important component of a pattern recognition system. It performs two tasks: transforming input parameter vector into a feature vector and/or reducing its dimensionality. A well-defined feature extraction algorithm makes the classification process more effective and efficient. Two popular methods for feature extraction are linear discriminant analysis (LDA) and principal component analysis (PCA). In this paper, the minimum classification error (MCE) training algorithm (which was originally proposed for optimizing classifiers) is investigated for feature extraction. A generalized MCE (GMCE) training algorithm is proposed to mend the shortcomings of the MCE training algorithm. LDA, PCA, and MCE and GMCE algorithms extract features through linear transformation. Support vector machine (SVM) is a recently developed pattern classification algorithm, which uses non-linear kernel functions to achieve non-linear decision boundaries in the parametric space. In this paper, SVM is also investigated and compared to linear feature extraction algorithms.

Fast features for face authentication under illumination direction changes

In this letter we propose a facial feature extraction technique which utilizes polynomial coefficients derived from 2D Discrete Cosine Transform (DCT) coefficients obtained from horizontally and vertically neighbouring blocks. Face authentication results on the VidTIMIT database suggest that the proposed feature set is superior (in terms of robustness to illumination changes and discrimination ability) to features extracted using four popular methods: Principal Component Analysis (PCA), PCA with histogram equalization pre-processing, 2D DCT and 2D Gabor wavelets; the results also suggest that histogram equalization pre-processing increases the error rate and offers no help against illumination changes. Moreover, the proposed feature set is over 80 times faster to compute than features based on Gabor wavelets. Further experiments on the Weizmann database also show that the proposed approach is more robust than 2D Gabor wavelets and 2D DCT coefficients.

Fast principal component analysis using fixed-point algorithm

In this paper we present an efficient way of computing principal component analysis (PCA). The algorithm finds the desired number of leading eigenvectors with a computational cost that is much less than that from the eigenvalue decomposition (EVD) based PCA method. The mean squared error generated by the proposed method is very similar to the EVD based PCA method.

Identity verification using speech and face information

This article first provides an overview of important concept s in the field of information fusion, followed by a review of milestones in audio-visual person identification and verifi cation. Several recent adaptive and non-adaptive techniques for reaching the verification decision (i.e., to accept or re ject the claimant), based on speech and face information, are then evaluated in clean and noisy audio conditions on a common database; it is shown that in clean conditions most of the non-adaptive approaches provide similar performance and in noisy conditions most exhibit a severe deterioration i n performance; it is also shown that current adaptive approac hes are either inadequate or utilize restrictive assumptio ns. A new category of classifiers is then introduced, where the de cision boundary is fixed but constructed to take into account how the distributions of opinions are likely to change due to noisy conditions; compared to a previously proposed adaptive approach, the proposed classifiers do not make a direct assum ption about the type of noise that causes the mismatch between training and testing conditions.

Gram-positive and Gram-negative protein subcellular localization by incorporating evolutionary-based descriptors into Chou׳s general PseAAC

Protein subcellular localization is defined as predicting the functioning location of a given protein in the cell. It is considered an important step towards protein function prediction and drug design. Recent studies have shown that relying on Gene Ontology (GO) for feature extraction can improve protein subcellular localization prediction performance. However, relying solely on GO, this problem remains unsolved. At the same time, the impact of other sources of features especially evolutionary-based features has not been explored adequately for this task. In this study, we aim to extract discriminative evolutionary features to tackle this problem. To do this, we propose two segmentation based feature extraction methods to explore potential local evolutionary-based information for Gram-positive and Gram-negative subcellular localizations. We will show that by applying a Support Vector Machine (SVM) classifier to our extracted features, we are able to enhance Gram-positive and Gram-negative subcellular localization prediction accuracies by up to 6.4% better than previous studies including the studies that used GO for feature extraction.

Fast K-dimensional tree algorithms for nearest neighbor search with application to vector quantization encoding

Fast search algorithms are proposed and studied for vector quantization encoding using the K-dimensional (K-d) tree structure. Here, the emphasis is on the optimal design of the K-d tree for efficient nearest neighbor search in multidimensional space under a bucket-Voronoi intersection search framework. Efficient optimization criteria and procedures are proposed for designing the K-d tree, for the case when the test data distribution is available (as in vector quantization application in the form of training data) as well as for the case when the test data distribution is not available and only the Voronoi intersection information is to be used. The criteria and bucket-Voronoi intersection search procedure are studied in the context of vector quantization encoding of speech waveform. They are empirically observed to achieve constant search complexity for O(log N) tree depths and are found to be more efficient in reducing the search complexity. A geometric interpretation is given for the maximum product criterion, explaining reasons for its inefficiency with respect to the optimization criteria. >

Improving prediction of secondary structure, local backbone angles, and solvent accessible surface area of proteins by iterative deep learning

Direct prediction of protein structure from sequence is a challenging problem. An effective approach is to break it up into independent sub-problems. These sub-problems such as prediction of protein secondary structure can then be solved independently. In a previous study, we found that an iterative use of predicted secondary structure and backbone torsion angles can further improve secondary structure and torsion angle prediction. In this study, we expand the iterative features to include solvent accessible surface area and backbone angles and dihedrals based on Cα atoms. By using a deep learning neural network in three iterations, we achieved 82% accuracy for secondary structure prediction, 0.76 for the correlation coefficient between predicted and actual solvent accessible surface area, 19° and 30° for mean absolute errors of backbone φ and ψ angles, respectively, and 8° and 32° for mean absolute errors of Cα-based θ and τ angles, respectively, for an independent test dataset of 1199 proteins. The accuracy of the method is slightly lower for 72 CASP 11 targets but much higher than those of model structures from current state-of-the-art techniques. This suggests the potentially beneficial use of these predicted properties for model assessment and ranking.