论文部分内容阅读
1 引言在汉语文-语转换系统(CTTS)的研究中,要求系统合成的语音应尽量模拟自然语音,体现出语言抑扬顿挫的美感。这需要正确地界定韵律短语,细微地模拟自然语音的韵律模式。自然语音在字调、语调、重音等方面的变化,其声学特征的表现为音长、基频和幅度等时域参数。因此,在基于时域PSOLA的CTTS研究
1 Introduction In the study of Chinese-to-language conversion system (CTTS), it is required that the synthesized speech should try to simulate the natural speech as much as possible to reflect the beauty of the language. This requires proper definition of prosodic phrases that subtly simulate the prosodic patterns of natural speech. The variation of natural speech in tone, intonation, accent and so on, the performance of its acoustic features are time-domain parameters such as pitch, fundamental frequency and amplitude. Therefore, a CTTS study based on time-domain PSOLA