連続音声中の母音連鎖における調音結合効果の正規化

スポンサーリンク

概要

• 論文の詳細を見る

• In our previous study, it was found that the phonetic values of vowels taken out from a connected speech were perceptually quite different from the corresponding isolated monosyllables, and that two syllables, one preceding and one following, were necessary to provide a perceptual environment for their correct identification. These facts indicate that the perception of vowels in connected speech is seriously impaired by the complete removal of their environment. In other words, the identity of such a vowel is not only determined by locally extracted features but also by its relationship with those of the preceding and following sounds. This paper is concerned with the dynamic aspect of vowel reduction, associated with contextual influence, and the modification method by which such vowels are characterized in acoustical features using contextual information. It is assumed that the acoustical features of speech sound at any instant, are modified by the contribution of dynamic acoustical features of preceding and succeeding continuous signals, and that the extent of the effect conforms to the normal distribution function of time(Eq. (5)). Two unknown factors contained in this function were calculated from the results of the perceptual experiment by B. E. Lindblom, and the modification rule was applied to the middle vowels in /VV'V/ context(symmetric form)and in /VV'V''/ context(asymmetric form). The instantaneous first and second formant frequencies of vowels were used as the acoustical features and the modification was done obeying Eq. (8). The results were as follows:1. Normalization effect was observed for vowels in the case of the symmetric form. 2. Each vowel region in the f1-f2 diagram shifted towards the center of the distribution for all vowels as compared with the region of monosyllabic vowels and partially overlapped with each other, but vowel regions were separably located in modified f1-f2 diagram.

• 社団法人日本音響学会の論文
• 1973-02-01

著者

• 桑原 尚夫

  帝京科学大学

• 境 久雄

  日本放送協会放送科学基礎研究所

• 桑原 尚夫

  日本放送協会放送科学基礎研究所

関連論文

• 研究用日本語音声データベースの構築
• ベクトル量子化による声質変換
• 国内における音声データベースの現状 : 開発,管理及び音声研究への利用
• 音声研究の40年 : 音声知覚、音声個人性、声質変換、音声データベースの研究(若手研究者育成レクチャーシリーズ)(音声の基礎と応用シンポジウム)
• XML 教材構築システムの作成
• 波形処理による鼻濁音の分析
• ホルマント周波数の時間変化パターンと2連および3連母音の知覚
• 合成音声の両耳融合と動的母音の知覚機構
• ホルマント周波数の種々の時間変化に対する母音の知覚
• 動的合成母音による音韻境界の知覚的検討
• 連続音声中の母音連鎖における調音結合効果の正規化
• 連続音声中の切り出し母音および音節の音韻知覚
• 耳の感受性と音のうるささ
• 5. [若手研究者育成レクチャーシリーズ]音声研究の40年 : 音声知覚,音声個人性,声質変換,音声データベースの研究(第309回研究例会発表要旨)

もっと見る 閉じる

スポンサーリンク