Wu Chou

Bio: Wu Chou is an academic researcher from Chinese Academy of Sciences. The author has contributed to research in topics: Linear regression & Pattern recognition (psychology). The author has an hindex of 6, co-authored 7 publications receiving 279 citations. Previous affiliations of Wu Chou include Avaya & Indian Institute of Technology Madras.

Discriminative learning in sequential pattern recognition

Abstract: In this article, we studied the objective functions of MMI, MCE, and MPE/MWE for discriminative learning in sequential pattern recognition. We presented an approach that unifies the objective functions of MMI, MCE, and MPE/MWE in a common rational-function form of (25). The exact structure of the rational-function form for each discriminative criterion was derived and studied. While the rational-function form of MMI has been known in the past, we provided the theoretical proof that the similar rational-function form exists for the objective functions of MCE and MPE/MWE. Moreover, we showed that the rational function forms for objective functions of MMI, MCE, and MPE/MWE differ in the constant weighting factors CDT (s1 . . . sR) and these weighting factors depend only on the labeled sequence s1 . . . sR, and are independent of the parameter set - to be optimized. The derived rational-function form for MMI, MCE, and MPE/MWE allows the GT/EBW-based parameter optimization framework to be applied directly in discriminative learning. In the past, lack of the appropriate rational-function form was a difficulty for MCE and MPE/MWE, because without this form, the GT/EBW-based parameter optimization framework cannot be directly applied. Based on the unified rational-function form, in a tutorial style, we derived the GT/EBW-based parameter optimization formulas for both discrete HMMs and CDHMMs in discriminative learning using MMI, MCE, and MPE/MWE criteria. The unifying review provided in this article has been based upon a large number of earlier contributions that have been cited and discussed throughout the article. Here we provide a brief summary of such background work. Extension to large-scale speech recognition tasks was accomplished in the work of [59] and [60]. The dissertation of [47] further improved the MMI criterion to that of MPE/MWE. In a parallel vein, the work of [20] provided an alternative approach to that of [41], with an attempt to more rigorously provide a CDHMM model re-estimation formula that gives positive growth of the MMI objective function. A crucial error of this attempt was corrected in [2] for establishing an existence proof of such positive growth. The main goal of this article is to provide an underlying foundation for MMI, MCE, and MPE/MWE at the objective function level to facilitate the development of new parameter optimization techniques and to incorporate other pattern recognition concepts, e.g., discriminative margins [66], into the current discriminative learning paradigm.

Minimum classification error linear regression for acoustic model adaptation of continuous density HMMs

Abstract: In this paper, a concatenated "super" string model based minimum classification error (MCE) model adaptation approach is described. We show that the error rate minimization in the proposed approach can be formulated into maximizing a special ratio of two positive functions. The proposed string model is used to derive the growth transform based error rate minimization for MCE linear regression (MCELR). It provides an effective solution to apply MCE approach to acoustic model adaptation with sparse data. The proposed MCELR approach is studied and compared with the maximum likelihood linear regression (MLLR) based model adaptation. Experiments on large vocabulary speech recognition tasks are performed. Experimental results indicate that the proposed MCELR model adaptation can lead to significant speech recognition performance improvement and its performance advantage over the MLLR based approach is observed even when the amount of adaptation data is sparse.

A Novel Learning Method for Hidden Markov Models in Speech and Audio Processing

Abstract: in recent years, various discriminative learning techniques for HMMs have consistently yielded significant benefits in speech recognition. In this paper, we present a novel optimization technique using the Minimum Classification Error (MCE) criterion to optimize the HMM parameters. Unlike Maximum Mutual Information training where an Extended Baum-Welch (EBW) algorithm exists to optimize its objective function, for MCE training the original EBW algorithm cannot be directly applied. In this work, we extend the original EBW algorithm and derive a novel method for MCE-based model parameter estimation. Compared with conventional gradient descent methods for MCE learning, the proposed method gives a solid theoretical basis, stable convergence, and it is well suited for the large-scale batch-mode training process essential in large-scale speech recognition and other pattern recognition applications. Evaluation experiments, including model training and speech recognition, are reported on both a small vocabulary task (TI-Digits) and a large vocabulary task (WSJ), where the effectiveness of the proposed method is demonstrated. We expect new future applications and success of this novel learning method in general pattern recognition and multimedia processing, in addition to speech and audio processing applications we present in this paper.

Minimum classification error (MCE) model adaptation of continuous density HMMS.

Abstract: In this paper, a framework of minimum classification error (MCE) model adaptation for continuous density HMMs is proposed based on the approach of "super" string model. We show that the error rate minimization in the proposed approach can be formulated into maximizing a special ratio of two positive functions, and from that a general growth transform algorithm is derived for MCE based model adaptation. This algorithm departs from the generalized probability descent (GPD) algorithm, and it is well suited for model adaptation with a small amount of training data. The proposed approach is applied to linear regression based variance adaptation, and the close form solution for variance adaptation using MCE linear regression (MCELR) is derived. The MCELR approach is evaluated on large vocabulary speech recognition tasks. The relative performance gain is more than doubled on the standard (WSJ Spoke 3) database, comparing to maximum likelihood linear regression (MLLR) based variance adaptation for the same amount of adaptation data.

Maximum a posteriori linear regression (MAPLR) variance adaptation for continuous density HMMS.

Abstract: In this paper, the theoretical framework of maximum a posteriori linear regression (MAPLR) based variance adaptation for continuous density HMMs is described. In our approach, a class of informative prior distribution for MAPLR based variance adaptation is identified, from which the close form solution of MAPLR based variance adaptation is obtained under its EM formulation. Effects of the proposed prior distribution in MAPLR based variance adaptation are characterized and compared with conventional maximum likelihood linear regression (MLLR) based variance adaptation. These findings provide a consistent Bayesian theoretical framework to incorporate prior knowledge in linear regression based variance adaptation. Experiments on large vocabulary speech recognition tasks were performed. The experimental results indicate that significant performance gain over the MLLR based variance adaptation can be obtained based on the proposed approach.

Deep Neural Networks for Acoustic Modeling in Speech Recognition: The Shared Views of Four Research Groups

Abstract: Most current speech recognition systems use hidden Markov models (HMMs) to deal with the temporal variability of speech and Gaussian mixture models (GMMs) to determine how well each state of each HMM fits a frame or a short window of frames of coefficients that represents the acoustic input. An alternative way to evaluate the fit is to use a feed-forward neural network that takes several frames of coefficients as input and produces posterior probabilities over HMM states as output. Deep neural networks (DNNs) that have many hidden layers and are trained using new methods have been shown to outperform GMMs on a variety of speech recognition benchmarks, sometimes by a large margin. This article provides an overview of this progress and represents the shared views of four research groups that have had recent successes in using DNNs for acoustic modeling in speech recognition.

Context-Dependent Pre-Trained Deep Neural Networks for Large-Vocabulary Speech Recognition

Abstract: We propose a novel context-dependent (CD) model for large-vocabulary speech recognition (LVSR) that leverages recent advances in using deep belief networks for phone recognition. We describe a pre-trained deep neural network hidden Markov model (DNN-HMM) hybrid architecture that trains the DNN to produce a distribution over senones (tied triphone states) as its output. The deep belief network pre-training algorithm is a robust and often helpful way to initialize deep neural networks generatively that can aid in optimization and reduce generalization error. We illustrate the key components of our model, describe the procedure for applying CD-DNN-HMMs to LVSR, and analyze the effects of various modeling choices on performance. Experiments on a challenging business search dataset demonstrate that CD-DNN-HMMs can significantly outperform the conventional context-dependent Gaussian mixture model (GMM)-HMMs, with an absolute sentence accuracy improvement of 5.8% and 9.2% (or relative error reduction of 16.0% and 23.2%) over the CD-GMM-HMMs trained using the minimum phone error rate (MPE) and maximum-likelihood (ML) criteria, respectively.

Deep Learning: Methods and Applications

Abstract: This monograph provides an overview of general deep learning methodology and its applications to a variety of signal and information processing tasks. The application areas are chosen with the following three criteria in mind: (1) expertise or knowledge of the authors; (2) the application areas that have already been transformed by the successful use of deep learning technology, such as speech recognition and computer vision; and (3) the application areas that have the potential to be impacted significantly by deep learning and that have been experiencing research growth, including natural language and text processing, information retrieval, and multimodal information processing empowered by multi-task deep learning.

Deep Neural Networks for Acoustic Modeling in Speech Recognition

Abstract: Most current speech recognition systems use hidden Markov models (HMMs) to deal with the temporal variability of speech and Gaussian mixture models (GMMs) to determine how well each state of each HMM fits a frame or a short window of frames of coefficients that represents the acoustic input. An alternative way to evaluate the fit is to use a feed-forward neural network that takes several frames of coefficients as input and produces posterior probabilities over HMM states as output. Deep neural networks (DNNs) that have many hidden layers and are trained using new methods have been shown to outperform GMMs on a variety of speech recognition benchmarks, sometimes by a large margin. This article provides an overview of this progress and represents the shared views of four research groups that have had recent successes in using DNNs for acoustic modeling in speech recognition.

Convolutional neural networks for speech recognition

Abstract: Recently, the hybrid deep neural network (DNN)- hidden Markov model (HMM) has been shown to significantly improve speech recognition performance over the conventional Gaussian mixture model (GMM)-HMM. The performance improvement is partially attributed to the ability of the DNN to model complex correlations in speech features. In this paper, we show that further error rate reduction can be obtained by using convolutional neural networks (CNNs). We first present a concise description of the basic CNN and explain how it can be used for speech recognition. We further propose a limited-weight-sharing scheme that can better model speech features. The special structure such as local connectivity, weight sharing, and pooling in CNNs exhibits some degree of invariance to small shifts of speech features along the frequency axis, which is important to deal with speaker and environment variations. Experimental results show that CNNs reduce the error rate by 6%-10% compared with DNNs on the TIMIT phone recognition and the voice search large vocabulary speech recognition tasks.