Speech pitch determination based on Hilbert-Huang transform

Pitch determination is an essential part of speech recognition and speech processing. In this paper, a new pitch determination method based on Hilbert-Huang Transform (HHT) is presented. The assumption of linearity of the speech-production process and short-time stationarity of speech signals, which is generally employed in recent studies on speech recognition, is no longer needed, and hence the non-linearity of the speech-production process and the non-stationarity of speech signals shown in pitch characteristics can be well represented in processing of speech signals with this method. The YOHO speech database is applied in this work for the validation of the pitch determination method. The results show that, compared with conventional methods, the pitch determination method based on HHT well improves the accuracy and resolution of pitch recognition. (c) 2005 Elsevier B.V. All rights reserved.