Epoch extraction of voiced speech

Abstract

A general theory of epoch extraction of overlapping nonidentical waveforms is presented. The theory is applied to outputs of models of voiced speech production mechanism and to actual speech data. Some typical glottal waveshapes are considered to explain their effect on the speech output. It is shown that the points of excitation of the vocal tract can be precisely identified for continuous speech. It is possible to obtain accurate pitch information by this method even for high-pitched sounds. The epoch extraction has wide applications in speech analysis, speaker verification, speech synthesis, and pitch perception studies.