Speech recording for the man-computer communication in room with reverberation and cocktail party interference is serious problem. To improve the speech quality, microphone arrays are often used. Adaptive algorithms that exploits minimum variance (MV) criterion, are inefficient in interference suppression when there is any correlation between interference and desired signal. This correlation causes cancellation of the desired signal and signal-tointerference ratio reduction. The desired signal cancellation can be prevented if adaptive weightings are estimated in pause of speech. In this paper the subbands pause detection is proposed and applied to GSC weightings estimation. Pause detection is based on signal-to-interference ratio estimation. Unknown probability density functions of testing random variable are estimated from the histogram of mixture of hypothesis H0 (pause) and H1 (speech). The optimal pause detection threshold determined from estimated density functions guarantees low level of desired signal cancellation. The proposed pause detection algorithm is tested on simulation of the room with reverberation.
Cite as: Saric, Z.M., Jovicic, S.T. (2004) Subband pause in speech signal detection using microphone array in room with reverberation. Proc. 9th Conference on Speech and Computer (SPECOM 2004), 132-137
@inproceedings{saric04_specom,
author={Zoran M. Saric and Slobodan T. Jovicic},
title={{Subband pause in speech signal detection using microphone array in room with reverberation}},
year=2004,
booktitle={Proc. 9th Conference on Speech and Computer (SPECOM 2004)},
pages={132--137}
}