We have studied the analysis and synthesis of speech with a glottal-excited speech synthesizer. Instead of delta pulses, as used in a standard LPC synthesizer, this synthesizer uses glottal pulses as excitation funclion. In the present paper we discuss how the parameters of the new speech synthesizer are derived from the speech signal. First we estimate the vocal-tract filter pitch synchronously. The glottal waveform is then derived by inverse filtering the speech signal. Third, the glottal waveform is stylized using the four-parameter model of Liljencrants and Fant (Fant el al., 1985).
|Journal||IPO Annual Progress Report|
|Publication status||Published - 1989|