A new method for extracting the glottal closure intervals of voiced speech

概要

論文の詳細を見る
The purpose of the paper is to find out the instant of the glottal opening/closure from the speech signal by a recursive algorithm, and to accurately estimate the characteristics of a vocal tract by using glottal closure covariance analysis. The identification of the glottal closure/opening was achieved by estimating the waveform of the glottal volume flow. A recursive estimation of the glottal volume flow was realized by using a time-varying AR model with an unknown nonwhite input (u-input). In the applications to synthetic vowels generated by the two-mass model, it is shown that the phase of the estimated u-input corresponds to that of the glottal volume flow. It is also shown in the experiment for real vowels uttered by male adults that the phase of the estimated u-input crresponds to that of the glottal movements which can be confirmed by observing the Electroglottograph (EGG). The glottal closure intervals are determined from the estimated u-input, and stable formant frequencies are estimated by covariance analysis which utilizes only the speech during glottal closure intervals. The estimates are compared to the conventional autocorrelation LP analysis. As results of some experiments, we show that the proposed method is effective for the identification of the glottal closure intervals
社団法人日本音響学会の論文