Estimation of articulatory motion by a model matching method
Katsuhiko Shirai, Masayuki Honda · The Journal of the Acoustical Society of America · 1978
First, an effective method to describe the position of the articulatory organs is presented. The method relies on the statistical analysis of real data. The vocal tract shape is specified effectively by five articulatory parameters, the angle of jaw opening, two principal components for the various tongue shapes, lip parameter, and glottal parameter. Next, this articulatory model is used to estimate articulatory motion from speech waves. The articulatory parameters are estimated by minimizing the spectral error with some constraints. This is a nonlinear optimization problem and is solved by an iterative procedure. The power spectrum of the speech wave is composed of the vocal tract transfer function and the glottal wave characteristic. The spectral shape of the vocal tract transfer function is separated in such a way that the formant structure is preserved. Combined with a procedure to adjust the mean vocal tract length for each speaker, this method is very successful in estimating the articulatory parameters from vowel sounds. [Work supported by Ministry of Education Grant-in-Aid for Scientific Research No. 239005]