Determination of the Vocal-Tract Shape Directly from the Speech Wave

Publisher Website
Google Scholar
Abstract
Sound propagation through the human vocal tract is studied by quantizing the vocal tract into a finite but arbitrarily large number of contiguous uniform cylindrical sections. The transfer function between volume velocities at the glottis and lips and the corresponding impulse response are obtained by matching the pressure and the volume velocity at the junctions between adjacent sections. It is shown that the frequencies and bandwidths of the poles of the transfer function determine uniquely the areas of the cylindrical sections. The vocal-tract area function is determined directly from the autocorrelation function of the impulse response and the calculation does not involve any iterative or search procedure. The application of this procedure to determine the vocal-tract shape for connected speech is discussed.