Qiguang Lin
CAIP Ctr., Rutgers Univ., Piscataway, NJ 08855-1390
This paper studies a three-channel model of the vocal tract for nasals and nasalization. Traditionally, a two-channel model has been utilized: One channel for pharyngeal-oral tract and one for the nasal tract. This representation assumes that the left and right branches of the anterior part of the nasal tract are symmetrical and that the corresponding paranasal cavities connected to either branch are also symmetrical. Recently available data of the nasal geometry show, however, an apparent asymmetry between the left and right branches and between the companion sinus cavities [Dang et al., J. Acoust. Soc. Am. 94, 1765(A) (1993)]. To take into account the differences, the nasal tract must be simulated by a two-channel model, resulting in a three-channel model for the whole vocal system. Based on the proposed three-channel model, calculations are made of spectral patterns of nasals and nasalization. The results show a bound pole-zero structure around 2000 Hz. This pole-zero combination stems from the asymmetry between the left and right branches, and conforms with sweep-tone measurements. Other differences from conventional modeling of the vocal tract are also noted. [The simulation program is available upon request.]