It is important to investigate articulatory motion to obtain fundamental knowledge in speech communication, speech synthesis and speech recognition. The characteristics of articulatory motion should thus be reflected in data processing in the acoustic domain. For this purpose it is desirable to establish a suitable model of the articulatory mechanism. This paper introduces a new method of constructing such a model. The special feature of this model is representation of the variation of the tongue shape using principal components obtained by the analysis of X-ray data. This permits more accurate description of the articulation than the usual models using the same number of parameters. Correspondences between the articulatory parameters and the acoustic features are calculated in this model and the results are applied for estimation of the articulatory parameters by a nonlinear regression method. Results obtained by estimation from the formant frequencies are successfully applied to synthesized and real voices and show the validity of this method.
|ジャーナル||Electron Commun Jpn|
|出版物ステータス||Published - 1976 1 1|
ASJC Scopus subject areas