Vocal tract modeling: implementation of continuous length variations in a half-sample delay Kelly-Lochbaum model

Algorithms for modeling static vowel shapes using a half-sample delay Kelly-Lochbaum model are described, so that the total tract length is no longer limited to an integer multiple of the segment length, but can take on any continuous value. Fractional-delay filters are used for modeling lengthening of lips, lowering of the larynx and lengthening of an intermediate segment. A realistic vowel shape with no losses is modeled as a test case. It has 44 cylindrical segments and operates at a sampling frequency of 44.1 KHz.