Third International Conference on Spoken Language Processing (ICSLP 94)

Yokohama, Japan
September 18-22, 1994

Possibility of Speech Synthesis by Common Voice Source

Satoshi Yumoto, Jouji Suzuki, Tetsuya Shimamura

Dept.of Information and Computer Sciences, Saitama University, Urawa, Japan

In order to improve naturalness of synthesized speech, this paper intend to utilize common residual wave as voice source of Japanese typical sentences which have the same intonation or accent pattern. An analysis-synthesis with 14 coefficients are employed in this experiment. The synthetic experiments are conducted on some words. At first, a residual wave extracted from a word, are used as excitation source waves for other words. Secondly, fundamental frequency and intensity extracted by the original words are used to generate source waves. This synthesized sounds are less noisy than aforesaid sounds. In the last, fundamental frequency is generated by Fujisaki-model. This paper suggests that there are some possibilities to use common voice source for synthesis.

Full Paper

Bibliographic reference.  Yumoto, Satoshi / Suzuki, Jouji / Shimamura, Tetsuya (1994): "Possibility of speech synthesis by common voice source", In ICSLP-1994, 1675-1678.