In this paper, acoustic properties of realistic vocal tract models for sibilant fricatives /s/, /C/ and /∫/ derived from MRI data are calculated by means of the finite element method (FEM). The FEM simulation reproduced relative frequency distribution among the sibilants (/s/ >/c/ >/∫/), and detailed spectral structure coincided with the spectral pattern of recorded sounds. The pressure distribution inside the vocal tract models examined at the frequency of poles permit to explain the origin of spectral preeminence characteristic of the sibilants: several resonances were clustered in these frequencies. As for /s/, resonances of the vocal tract behind the constriction was responsible of this concentration. /C/ and /∫/ showed a different pattern: resonances in the anterior parts of the vocal tract (e.g. sublingual cavity and lateral interdental space) were situated in the vicinity of one of the resonances of the vocal tract behind the constriction. The results suggest that the sibilants' acoustic target is determined by the possible acoustic patterns where several resonances come close together so that they reinforce with each other for efficient acoustic realization.
展开▼