Speech Synthesis Algorithm of Co-articulation Based on the Continuous Transition of Energy
Guangyu Kang, Shize Guo, Longjiang Yu
Abstract
Guangyu Kang, Shize Guo, Longjiang Yu
Abstract
Co-articulation is the phenomenon of co-occurrence of two or more than two phonemes features affected by front and behind phonemes during articulation. In the current waveform concatenation synthesis methods in Chinese speech synthesis, the effect of behind syllable on the front syllable is not considered, which results in co-articulation inefficient synthesis in the joint of syllables. The co-articulation inefficient synthesis will decrease the naturalness of the synthesized speech. To solve this problem, an algorithm is proposed in this paper to perform co-articulation synthesis by modification of the speech transitional energy spectrum based on the research of the phenomenon of the co-articulation in Chinese speech. The naturalness of speech synthesis by the proposed method in this paper is higher and lower demand on the tone database by experimental results.
A significance statement is not available in the OpenAlex record.
A contribution statement is not available in the OpenAlex record.
Method details are not available in the OpenAlex metadata.
Findings are not separately available in the OpenAlex metadata.
Limitations are not available in the OpenAlex metadata.
Application details are not available in the OpenAlex metadata.
Co-articulation is the phenomenon of co-occurrence of two or more than two phonemes features affected by front and behind phonemes during articulation. In the current waveform concatenation synthesis methods in Chinese speech synthesis, the effect of behind syllable on the front syllable is not considered, which results in co-articulation inefficient synthesis in the joint of syllables. The co-articulation inefficient synthesis will decrease the naturalness of the synthesized speech. To solve this problem, an algorithm is proposed in this paper to perform co-articulation synthesis by modification of the speech transitional energy spectrum based on the research of the phenomenon of the co-articulation in Chinese speech. The naturalness of speech synthesis by the proposed method in this paper is higher and lower demand on the tone database by experimental results.
Key concepts: Naturalness, Articulation (sociology), Concatenation (mathematics), Speech synthesis, Speech recognition, Speech production, Computer science, Coarticulation