Third European Conference on Speech Communication and Technology

Berlin, Germany
September 22-25, 1993


A Dynamic Approach to Speaker Adaptation of Hidden Markov Networks for Speech Recognition

Tetsuo Kosaka (1), Edward Willems (2), Jun-Ichi Takami (1), Shigeki Sagayama (3)

(1) ATR Interpreting Telecommunications Research Labs., Kyoto, Japan
(2) Ecole Nationale Superieure des Telecommunications, Paris, France
(3) 3NTT Human Interface Research Labs., Tokyo, Japan

This paper describes a new approach to dynamic speaker adaptation, which relies on switching between different methods of adaptation in order to gain maximum performance depending on the amount of speech data obtained through the speech recognition session. This adaptaion method has been successfully applied to a hidden Markov network (HMnet), which is an efficient representation of phoneme context-dependent HMMs. This speaker adaptation method has proven itself effective in improving the performance of a Japanese phrase recognition system.

Full Paper

Bibliographic reference.  Kosaka, Tetsuo / Willems, Edward / Takami, Jun-Ichi / Sagayama, Shigeki (1993): "A dynamic approach to speaker adaptation of hidden Markov networks for speech recognition", In EUROSPEECH'93, 363-366.