Speaker adaptation through spectral transformation for HMM based speech recognition

Author

Choi, H.C. ; King, R.W.

Author_Institution

Dept. of Electr. Eng., Sydney Univ., NSW, Australia

fYear

1994

fDate

13-16 Apr 1994

Firstpage

686

Abstract

The use of spectral transformation to perform speaker adaptation for HMM based speech recognition is investigated. Three estimation methods, namely, minimum mean square error (MMSE), canonical correlation analysis (CCA) and multilayer perceptrons (MLP), for computing the transformation are compared. Using isolated words from the TI-46 database, it is found that CCA has the best adaptation performance. Moreover, a “training-after-adaptation” approach is found to have a higher adaptation performance than the one in which reference HMMs are not re-trained. With a suitable choice of reference speaker, less than 30% of training data from a new speaker is required in order to achieve the same accuracy as the speaker-dependent models of that new speaker, when the CCA method is used with the “training-after-adaptation” approach

Keywords

correlation methods; error analysis; feedforward neural nets; hidden Markov models; spectral analysis; speech analysis and processing; speech recognition; HMM; MLP; MMSE; TI-46 database; adaptation performance; canonical correlation analysis; estimation methods; isolated words; minimum mean square error; multilayer perceptrons; reference speaker; speaker adaptation; speaker-dependent models; spectral transformation; speech recognition; training data; training-after-adaptation; Estimation error; Hidden Markov models; Mean square error methods; Multilayer perceptrons; Spatial databases; Speech recognition; Training data;

fLanguage

English

Publisher

ieee

Conference_Titel

Speech, Image Processing and Neural Networks, 1994. Proceedings, ISSIPNN '94., 1994 International Symposium on

Print_ISBN

0-7803-1865-X

Type

conf

DOI

10.1109/SIPNN.1994.344819

Filename

344819