NON-LINEAR NOISE COMPENSATION FOR ROBUST SPEECH RECOGNITION USING GAUSS-NEWTON METHOD

被引：0

作者：

Zhao, Yong ^{[1
]}

Juang, Biing-Hwang ^{[1
]}

机构：

[1] Georgia Inst Technol, Ctr Signal & Image Proc, Atlanta, GA 30332 USA

来源：

2011 IEEE INTERNATIONAL CONFERENCE ON ACOUSTICS, SPEECH, AND SIGNAL PROCESSING | 2011年

关键词：

Gauss-Newton method; non-linear compensation; robust speech recognition; vector Taylor series;

D O I：

暂无

中图分类号：

O42 [声学];

学科分类号：

070206 ; 082403 ;

摘要：

In this paper, we present the Gauss-Newton method as a unified approach to optimizing non-linear noise compensation models, such as vector Taylor series (VTS), data-driven parallel model combination (DPMC), and unscented transform (UT). We demonstrate that the commonly used approaches that iteratively approximate the noise parameters in an EM framework are variants of the Gauss-Newton method. Through the formulation of the Gauss-Newton method for estimating noise means and variances, the noise estimation problems are reduced to determining the Jacobians of the noisy speech distributions. For the sampling-based compensations, we present two methods, sample Jacobian average (SJA) and cross-covariance (XCOV), to evaluate the Jacobians. Experiments on the Aurora 2 database verify the efficacy of the Gauss-Newton method to these noise compensation models.

引用

页码：4796 / 4799

页数：4

共 38 条

[1] Nonlinear Compensation Using the Gauss-Newton Method for Noise-Robust Speech Recognition [J].

Zhao, Yong ;

Juang, Biing-Hwang .

IEEE TRANSACTIONS ON AUDIO SPEECH AND LANGUAGE PROCESSING, 2012, 20 (08) :2191-2206

[2] Phaseless Recovery Using the Gauss-Newton Method [J].

Gao, Bing ;

Xu, Zhiqiang .

IEEE TRANSACTIONS ON SIGNAL PROCESSING, 2017, 65 (22) :5885-5896

[3] Beyond Linear Transforms: Efficient Non-linear Dynamic Adaptation for Noise Robust Speech Recognition [J].

Rennie, Steven J. ;

Dognin, Pierre L. .

INTERSPEECH 2008: 9TH ANNUAL CONFERENCE OF THE INTERNATIONAL SPEECH COMMUNICATION ASSOCIATION 2008, VOLS 1-5, 2008, :1305-1308

[4] A comparative study of noise estimation algorithms for nonlinear compensation in robust speech recognition [J].

Zhao, Yong ;

Juang, Biing-Hwang .

SPEECH COMMUNICATION, 2017, 89 :58-69

[5] ON THE SEMILOCAL CONVERGENCE OF THE GAUSS-NEWTON METHOD USING RECURRENT FUNCTIONS [J].

Argyros, Ioannis K. ;

Hilout, Said .

JOURNAL OF THE KOREAN SOCIETY OF MATHEMATICAL EDUCATION SERIES B-PURE AND APPLIED MATHEMATICS, 2010, 17 (04) :307-319

[6] Gauss-Newton method for solving linear inverse problems with neural network coders [J].

Scherzer, Otmar ;

Hofmann, Bernd ;

Nashed, Zuhair .

SAMPLING THEORY SIGNAL PROCESSING AND DATA ANALYSIS, 2023, 21 (02)

[7] FREQUENCY DOMAIN ELASTIC WAVEFORM INVERSION USING THE GAUSS-NEWTON METHOD [J].

Chung, Wookeen ;

Shin, Jungkyun ;

Bae, Ho Seuk ;

Yang, Dongwoo ;

Shin, Changsoo .

JOURNAL OF SEISMIC EXPLORATION, 2012, 21 (01) :29-48

[8] An Iterative Algorithm for Microwave Tomography Using Modified Gauss-Newton Method [J].

Kundu, A. K. ;

Bandyopadhyay, B. ;

Sanyal, S. .

4TH KUALA LUMPUR INTERNATIONAL CONFERENCE ON BIOMEDICAL ENGINEERING 2008, VOLS 1 AND 2, 2008, 21 (1-2) :511-+

[9] ON THE GAUSS-NEWTON METHOD FOR CONVEX OPTIMIZATION USING RESTRICTED CONVERGENCE DOMAINS [J].

Argyros, Ioannis K. ;

George, Santhosh .

JOURNAL OF NONLINEAR FUNCTIONAL ANALYSIS, 2016,

[10] Feature compensation based on independent noise estimation for robust speech recognition [J].

Lu, Yong ;

Lin, Han ;

Wu, Pingping ;

Chen, Yitao .

EURASIP JOURNAL ON AUDIO SPEECH AND MUSIC PROCESSING, 2021, 2021 (01)

← 1 2 3 4 →