A new iterative algorithm for estimating the glottal flow derivative of vowels

Abstract

In this paper, we model the voiced speech signal as an AR process with an AR filter whose coefficients are obtained using a new iterative model-based algorithm. In the proposed iterative algorithm the Liljencrants-Fant (LF) model of the glottal flow is fitted, at each iteration, to the glottal derivative waveform extracted by closed phase inverse filtering. Taking this signal as the desired output of an adaptive filter excited by speech, the inverse of the AR filter is calculated using a normalized LMS algorithm. The mean square error is consequently minimized between the resulting residual and the LF model. The next iteration begins by obtaining a new LF model to fit the residual signal obtained by filtering speech with the up-dated filter. Therefore, a new estimation of the glottal flow derivative waveform is obtained at each iteration. The algorithm stops when no considerable changes occur, in two consecutive iterations, in the glottal flow derivative. Finally, the glottal flow estimates of real voiced speech sounds /a/ and /e/ are given as examples.