A maximum likelihood procedure for a universal adaptation method based on HMM composition

Author

Minami, Yasuhiro ; Furui, Sadaki

Author_Institution

NTT Human Interface Labs., Tokyo, Japan

Volume

1

fYear

1995

fDate

9-12 May 1995

Firstpage

129

Abstract

This paper proposes an adaptation method for universal noise (additive noise and multiplicative distortion) based on the HMM composition (compensation) technique. Although the original HMM composition can be applied only to additive noise, our new method can estimate multiplicative distortion by maximizing the likelihood value. The signal-to-noise ratio is automatically estimated as part of the estimation of multiplicative distortion. Phoneme recognition experiments show that this method improves recognition accuracy for noisy and distorted speech

Keywords

adaptive signal processing; hidden Markov models; maximum likelihood estimation; noise; speech processing; speech recognition; HMM compensation; HMM composition; additive noise; distorted speech; maximum likelihood procedure; multiplicative distortion; noisy speech; phoneme recognition experiments; recognition accuracy; signal-to-noise ratio; speech processing; universal adaptation method; universal noise; Additive noise; Automatic speech recognition; Background noise; Distortion; Hidden Markov models; Humans; Laboratories; Maximum likelihood estimation; Noise figure; Speech enhancement; Speech recognition;

fLanguage

English

Publisher

ieee

Conference_Titel

Acoustics, Speech, and Signal Processing, 1995. ICASSP-95., 1995 International Conference on

Conference_Location

Detroit, MI

ISSN

1520-6149

Print_ISBN

0-7803-2431-5

Type

conf

DOI

10.1109/ICASSP.1995.479290

Filename

479290