荒井 隆行

軟らかいゲル素材を用いて，軟口蓋が動き鼻咽腔閉鎖をする声道の物理模型を製作した．その模型では，軟口蓋から咽頭壁に渡ってゲル素材を用い，鼻腔を含むその他の部分はアクリル素材を用いた． 4 つの鼻咽腔閉鎖パターン （Coronal, Circular, Circular with Passavant's ridge, Sagittal） を模擬するため，ゲル素材の軟口蓋部にひもを付けると共に，両脇と後方からアクリル棒を押し込むことで咽頭側壁と咽頭後壁を動かせるようにした．これらの動作によって鼻咽腔閉鎖の度合いを連続的に変化させた結果，鼻咽腔結合部の開存面積に応じて模型から生成される音声信号の開鼻声の程度が変化することを確認すると共に，スペクトル上に極零対などが現れることを確認した．We developed a physical model of the human vocal tract with a flexible velum using gel-type material. In this model, while the velum and pharyngeal wall were made of the gel-type material, the rest of the vocal-tract model including the nasal cavity was made of acrylic materials. We simulated four patterns of velopharyngeal closure (the Coronal, Circular, Circular with Passavanfs ridge, and Sagittal patterns), by raising the velum by means of a string attached to the velic portion while the posterior and lateral pharyngeal walls were pushed inward from the outside with acrylic bars. By continuously changing the degree of velopharyngeal closure while pulling the string, we confirmed the change of the degree of hypemasality of a speech signal produced from the model, and we confirmed acoustic evidences such as a pole-zero pair on its spectrum.

健常者４名（男性２名と女性２名）の/impee/発音時の４次元MRI撮像を行い、３つの鼻咽腔閉鎖パターンを認めた。軟口蓋、鼻咽腔開存部および軟口蓋の挙上に関連する口蓋帆挙筋（LVP筋）の安静時及び母音/i/と子音/p/発音時の位置と形状を、正中矢状断面と水平断面で観察した。その結果、安静時にLVP筋は楕円形状をしており、その長軸は前後方向を向いていたが、軟口蓋が最大挙上すると鼻咽腔開存部に向かって回転した（タイプA）また軟口蓋の中等度の挙上ではLVP筋の長軸は近心に平行移動した（タイプB)。(1)Coronal patternでは/i/と//p/発音時ともにタイプA、(2)Circular patternでは/p/発音時でタイプA、/i/発音時にタイプB、(3)Circular with Passavant's ridge patternでは/i/と/p/発音時、共にタイプBの運動がみられた。4D-MRI imagings were obtained from four normal subjects (two males and two females) during speech production /impee/. Three kinds of the velopharyngeal (VP) closure patterns were observed. From the observation of the movements of soft palate, velopharyngeal port and Levator Veli Palatini (LVP) muscle during rest, /i/ and /p/ production on both mid-sagittal plane and coronal plane, it was found that the shape and the position of the levator veli palatini muscle (LVP muscle) were changed according to the motion of the soft palate. Maximum movement of the soft palate resulted in the rotation toward to the velopharyngeal port of the major axis of LVP muscle (type A movement). Medium movement of soft palate resulted in the medial translation of LVP muscle (type B movement). Two subjects having Coronal VP closure pattern showed type A movement during both vowel /i/ and consonant /p/ production. The subject who having Circular VP closure pattern showed type A movement during /p/ production and type B movement during /i/ production. The subject having Circular with Passavant's ridge pattern showed type B movement both /i/ and /p/ production.

Rhythm can be viewed in two different ways. Acoustic approach views rhythm as the alternating pattern of high and low intensity, which is regarded as a syllable. Phonemic approach attributes rhythm to the phonemic complexity of syllable structure and calculates rhythm based on the durations of consonant and vowel intervals. This paper investigates how well the acoustic approach fits to the phonemic approach. It tests two algorithms adopted in our previous studies, which estimate syllable centers from intensity contours based on the calculation of RMS and correlation with a cosine curve. The results are evaluated against the criteria of the phonemic approach. It concludes that the algorithms are valid, hence syllable shapes can be captured from the intensity contour.

In speaker identification by listening, the identification rates vary depending on the speech contents presented to the subjects. It is reported that the nasals are more effective than the oral sounds for identifying speakers. The present study investigates the availability of the nasal sounds in terms of syllable structures. The results showed that the coda nasals are highly effective, though onset consonants are also important. As to the place of articulation, alveolar consonants in onset positions were more effective than bilabials, and the nasals were better than their oral counterparts were.

This study aimed to reduce ventilator noise in the speech signals of patients who use a ventilator. Two applications, spectral subtraction and adaptive filtering, were examined. For the first experiment, we employed a new technique based on spectral subtraction for ventilator noise reduction and then evaluated the result using mean opinion score. The results showed that the new technique was superior to the common approach of spectral subtraction in reducing ventilator noise. For the second experiment, an adaptive filter was used in a simulated environment to reduce ventilator noise and thr...