Dynamic analysis of Japanese vowels using articulatory model
Shun Ishizaki · The Journal of the Acoustical Society of America · 1982
The dynamic analysis starts from area functions obtained directly by analyzing speech waveforms, where the adaptive inverse filtering method [T. Nakajima et al., SCS Stockholm, 1–3 Aug. (1974)] is adopted. Applying the articulatory model [S. Ishizaki, Bull. Electrotech. Lab. 42, Nos. 11 & 12 (1978)] to the area functions, articulatory parameters such as the tongue position, the lip opening area and the vocal tract length are extracted. In continuous Japanese vowels' analysis, the parameters at the quasistationary part of the vowels are obtained. The parameters representing the articulatory organ movements are enhanced linearly to estimate target positions of the movements [S. Ishizaki, 6th IJCAI 1, 422–424 (1979)]. The enhanced articulatory positions are shown to move on toward the positions for each vowel which are obtained from isolated vowels. Such dynamic enhancement is examined in the formant domain and the result is shown to be corresponding with the articulatory one.