Estimation of vocal-tract area function from lip impulse response based on lossy vocal-tract digital filter model
Takemi Mochida, Masaaki Honda · The Journal of the Acoustical Society of America · 2001
A method for estimating vocal-tract area function from lip impulse response based on lossy vocal-tract digital filter model is provided. Previous methods of lip impulse response [M. M. Sondhi and J. R. Resnick, 73, 985–1002 (1983)] determines the area function from impulse response measured at lips by using the incident signal excited outside the oral cavity. This method is sensitive to the noise of measured impulse response. An alternative method [M. Honda and T. Mochida, Proc. 1st ETRW-SPM, 167–170 (1996)] is advantageous for accurate estimation under noisy acoustic condition, based on the vocal-tract digital filter model that optimizes the M sections of areas by using the impulse response longer than M samples. In this method, however, the vocal-tract model is based on the transmission line that treats the effect of loss in simplified form. The new method proposed in this study applies a mixture model of the lossless distributed circuit and the lumped circuit accounting for effects such as viscous loss, heat conduction loss, and loss due to wall vibration. The results on acoustic measurements have shown that the new method has better performance over the old method using optimization scheme based on simplified transmission line model. [Work supported by CREST, JST.]