A Theory on the Characteristics of Pitch and Formant Frequency Transitions in Human Speech
R. M. Doublier · The Journal of the Acoustical Society of America · 1973
Perceptual experiments have shown that the characteristics of fundamental frequency (pitch) and formant frequency variations usually represent a more important acoustic cue than those of the formant amplitudes. Unfortunately, however, the exact nature of these frequency transitions is far more subtle. If the speech production process is viewed as a closed-loop feedback control system, application of modern control theory techniques yields a rather interesting model for the vocal-tract control function. This system may be viewed in terms of the familiar phase-locked loop. The loop filter model, which appears to fit both theoretical considerations and the empirical (sound spectrograph) results, is the so-called “imperfect integrator,” with transfer function F(s) = (s+a)/(s+b). A discussion of this proposed model, including samples of synthetic speech, is presented.