Implementation and Effects of Single Channel Dereverberation Based on the Harmonic Structure of Speech

Speech signals recorded with a distant microphone usually contain reverberation, which degrades speech recognition performance. This is especially severe when the reverberation time exceeds 0.5 sec because the recognition performance cannot be improved sufficiently even if we use an acoustic model trained under matched reverberation conditions. Therefore, reverberation in speech signals should be removed prior to recognition. In this paper, we propose a single channel dereverberation method based on the harmonic structure of speech. Experimental results show that the present method can effectively suppress reverberation when the reverberation time is longer than 0.1 sec.