A PYRAMID-STRUCTURED WAVELET ALGOTITHM FOR DETECTING PITCH PERIOF OF SPEECH SIGNAL

Abstract

A new time-spectral domain algotithm based on the pyramid-structured wwavelet transform for detecting pitch period has been developed in this paper. The development of the pyramid-structured wavelet transform for pitch detection is motivated by the observation that the human speech signals can be modeled as quasi-periodic signals whose fundamental frequencies are located in 30~500Hz. The pyramid=structured wavelet transform is able to zoom ito dominant frequency (30~500Hz) channel, and reserces the fundamental characteristics of speech signals. Therefore, the proposed algotithm can efficiently determing the pitch period in the dominant frequency channel. From the various experiments, they show that the proposed scheme is not only suitable for wide-range pitch period of speech signals but also is robust to noise. The performance of the proposed scheme is compared with that of the time domain and the spectral domain pitch detectors and other presented wavelet based pitch detectors.