论文部分内容阅读
为提高语音合成系统的性能,产生自然流畅的合成语音,该文结合多种拼接点过渡平滑算法,提出了一种以语境相关的音素为基本单元的基于隐Markov(hidden Markov model,HMM)模型的英语拼接合成系统。该合成方法兼有拼接合成以及参数合成的优点,具有相对的灵活性,以及一定的语音自然度。以音素为基本单元尽可能减少了拼接点的个数,降低拼接失真。实验结果表明,多种平滑算法的采用,保证了拼接边界过渡平滑连贯,提高了最终的拼接效果。
In order to improve the performance of speech synthesis system and produce natural and smooth synthesized speech, this paper presents a hidden Markov model (HMM) based on context-dependent phonemes as the basic unit, Model of English splicing synthesis system. The synthesis method has the advantages of stitching synthesis and parameter synthesis, has relative flexibility, and certain natural degrees of speech. Using phonemes as the basic unit to reduce the number of splicing points as much as possible to reduce the splicing distortion. The experimental results show that the adoption of multiple smoothing algorithms ensures smooth and consistent splicing boundary transition and improves the final splicing effect.