Study on Robust Speech Recognition Based on Microphone Array and Probabilistic Signal Source Modelling Techniques

Masahide SUGIYAMA*, Hikaru DATE*, Michael COHEN*
and Kiyohiro SHIKANO**, Satoshi NAKAMURA**, Shiro ISE**

* School of Computer Science and Engineering The University of Aizu
Aizu-Wakamatsu, Fukushima 965-80
** Graduate school of Information Science Nara Institute of Science and Technology
8916-5 Takayama-machi, Ikoma-shi, Nara 630-01

In order to achieve a more natural and microphone-free speech dialogue system this report describes a study on robustness improvement techniques in the acoustic part and speech processing part. This report explains the following three components: speech-input system, noise control system and speech recognition system. The achievements of this year are as follows; a baseline evaluation system of speech recognition, preliminary evaluation of distance sensitive microphone speech input system, additive noise and transmission distortion HMM signal source composition, and evaluation of active noise control using impedance control.

Keywords: speech recognition, robustness improvement, probabilistic signal model, distance sensitive microphone system, active noise control