![]() |
International Workshop on Hands-Free Speech Communication (HSC2001)April 9-11, 2001 |
![]() |
This paper describes an efficient technique to improve noise robustness of speech recognition in car noisy environments by considering acoustic features of car interior noise. We analyzed the relation between the Articulation Index values and the recognition rates in car environments under various driving conditions. We clarified that the recognition rate significantly deteriorated when engine noise (periodic sound) components in the frequency range above 200 Hz were large. We developed a preprocessing technique to improve noise robustness under a large amount of engine noise conditions. This is a technique that the cutoff frequency of the front-end high pass filter is adaptively changed from 200 through 400 Hz according to the level of engine noise components. The average recognition rate of eight cars under an accelerating condition was improved by 11.9% by this technique, and the recognition rate of one of these cars was considerably improved by 38.6%.
Bibliographic reference. Hoshino, H. / Terashima, R. / Shimizu, T. / Wakita, T. (2001): "Improvement of noise robustness of speech recognition in car environments using acoustic features of car interior noise", In HSC2001, 151-154.