調音パラメータのホルマント周波数表現能力
スポンサーリンク
概要
- 論文の詳細を見る
The present study provides the result of an attempt to analyze statistically the performance of articulatory parameters to express formant frequencies. A quadratic curve is computationally fitted to the sampled points on the cineradiographic trace of the tongue profile to express the tongue position either on the coordinates relative to the mandible or relative to the maxilla. The tongue position is defined by the specific point in the quadratic curve on its axis just R_0 apart from the vertex, where R_0 is the mean radius of curvature at the vertex. The protrusions and aperture of the lips are reduced into a single parameter by the principal component analysis and used as an articulatory parameter. The mandible opening and the larynx height are also adopted as articulatory parameters. The first three formant frequencies are extracted by the LPC analysis from synchronously recorded speech sounds. Stepwise regression analysis of the formant frequencies is conducted on the various subsets of those articulatory parameters and the performances of the polynomial functions of those parameters to express the formant frequencies are examined. The result shows that the four polynomials of the articulatory parameters are effective to express the formant frequencies.
著者
-
橋本 清
The University Of Electro-communications
-
佐々木 邦彦
Tokyo Institute Of Technology (nagatsuda)
-
橋本 清
University of Electro-Communications
関連論文
- Canonical analysis concering musical melodies and the listeners' emotions
- 調音パラメータのホルマント周波数表現能力
- テンプレートユニットを持つニューラルネットワークによる日本語音素の認識
- 3次元舌モデルを用いた舌筋力の推定