8th International Conference on Spoken Language Processing

Jeju Island, Korea
October 4-8, 2004

Combination of Standard and Throat Microphones for Robust Speech Recognition in Highly Noisy Environments

Martin Graciarena, Federico Cesari, Horacio Franco, Greg Myers, Cregg Cowan, Victor Abrash

SRI International, USA

We present a method to combine the standard and throat microphone signals for noise-robust speech recognition. Our approach is to extend the probabilistic optimum filter algorithm to estimate the standard microphone clean speech feature vectors from both microphones' noisy speech feature vectors. We tested the proposed approach in two noisy speech recognition tasks. In the first task we used a large vocabulary continuous speech recognition system and noisy speech both using artificially added noise and recorded inside an M1 tank. In the second task we used a real-time system and noisy speech recorded in a highly noisy environment, a HMMWV vehicle. We used noise-canceling and throat microphones. Because of the highly adverse conditions in this second task we propose an extension of the combined microphone approach, which takes into account the level of noise captured by the throat microphone. The combined microphone approach outperforms the single microphone approach in all the recognition experiments.

Full Paper

Bibliographic reference.  Graciarena, Martin / Cesari, Federico / Franco, Horacio / Myers, Greg / Cowan, Cregg / Abrash, Victor (2004): "Combination of standard and throat microphones for robust speech recognition in highly noisy environments", In INTERSPEECH-2004, 809-812.