How does a mynah bird imitate human speech?

Publisher Website
Google Scholar
Abstract
Broad-band spectrograms and computer-generated spectra have been obtained of a trained Indian Hill mynah bird and its tutor each saying seven English sentences composed of four to 12 phonetic segments. Comparisons of sound spectrograms indicate that the bird is able to produce good approximations to virtually any spectral or temporal detail, although energy in frequencies below about 700 Hz (in the range of the first formant) is attenuated by about 10–20 dB relative to human spectra. Fundamental frequency (F0) contours, formant transitions, and turbulent noise spectra for fricatives and plosive bursts are especially well imitated. Contour shapes are better preserved than absolute formant or F0 values. Some vowel spectra have clear formant patterns but many have multiple resonances and formant splitting effects that cannot be accounted for in terms of an acoustic tube model of formant generation. A new description of the behavior of the mynah syrinx is proposed to account for the observed acoustic patterns. One of the paired external labia in the syrinx is implicated as the source of voicing energy (analogous to the human vocal cords). Some or all of the formant resonances are produced by exciting the mechanical resonant modes of the internal tympaniform membrane in the syrinx. The five pairs of intrinsic muscles that are associated with the syrinx are capable of controlling membrane tensions and regulating the position of the external labia in the syringeal airways during sound production.

This publication has 0 references indexed in Scilit: