音声波形からの声帯音源波形の観測

書誌事項

タイトル別名
  • Observation of Glottal Waveform from Speech Waves
  • オンセイ ハケイ カラ ノ セイタイ オンゲン ハケイ ノ カンソク

この論文をさがす

抄録

It is necessary for the study of vocal quality to know characteristics of a glottal source. Many methods for the direct or indirect observation of the glottal waveform or vibration of the vocal cords have been proposed. Most of them, however, are not suitable for the study of vocal quality. An indirect method, proposed by R. M. Miller, is considered to be applicable for this purpose, because it does not restrict phonation. His theory is based on inverse filtering of the transfer character of the vocal tract by means of simple analogue equipment. This paper proposes a more sophisticated method using digital techniques. Speech waves are transformed into a frequency domain in the first place and subsequent calculation is carried out in this domain. To estimate the transfer function of the vocal tract, "Analysis by Synthesis"technique is adopted. The source spectrum is calculated of digital filtering and converted into a time function or glottal waveform by inverse Fourier transform. This paper begins with the description of the theoretical background of the proposed method. Next, the effects of mismatching in the inverse filtering are examined with synthesized speech waves. An example of calculated glottal waves is shown in Fig. 1. Systematic evaluation of mismatching on every formant frequency and bandwidth is made with the results shown in Fig. 2 to 5. η in these figures is defined by Equation (10) and used for evaluation to denote waveform distortion caused by mismatching. Fig. 6 shows data handling procedure. In this experiment data must be treated carefully because of importance of waveform information. An experiment with natural speech is carried out according to the flow chart shown in Fig. 7. The most essential part of this experiment is the determination of the transfer function of the vocal tract. Modified "Analysis by Synthesis"technique is used for this purpose and the details are illustlated in Fig. 8. The glottal wave forms extracted from vowels uttered by several male speakers are drawn in Fig. 9 and 10. It is difficult to prove that these wave forms are true ones but they are considered to be acceptable as compared with the results of the preceding experiments. To get more precise results, the transfer function of the vocal tract must be determined more correctly. Further studies are required for solving this problem with due regard paid to the zero of a glottal source in the process of analysis by synthesis procedure. This method will be useful for the studies of naturalness and the individuality of speech sounds and of the mechanism of vocal cords vibration, and will contribute to the design of the excitation source of a synthesizer.

収録刊行物

  • 日本音響学会誌

    日本音響学会誌 26 (3), 141-149, 1970

    一般社団法人 日本音響学会

被引用文献 (2)*注記

もっと見る

詳細情報 詳細情報について

問題の指摘

ページトップへ