基于PSOLA算法的语音编码压缩的研究与实现

以时域基音同步叠加 (TD PSOLA)技术和一个全汉语单音节库为合成单元进行汉语语音合成 ,合成的语音清晰度和自然度很高 .但是这样的系统语音库太大 ,不利于在小型设备中实现 ,影响了语音合成的进一步应用 .本文针对此问题 ,在研究A律压缩的基础上 ,采用自适应量化和自适应预测的技术 ,以较少增加合成运算量复杂度为代价 ,对语音库的编码实现压缩 ,使压缩后的语音库减小了约一半 ,大大减小了所需的存储空间 .并且利用压缩后的语音库合成语音 ,基本上不影响合成后的语音质量 ,从而进一步扩展了语音合成的应用 .

Research and Realization of Coding Compress of Speech Synthesis Based on PSOLA

Based on the time domain pitch synchronous overlap add(PSOLA) method and a Chinese syllable dictionary Chinese speech waveform coding synthesis can be realized with high intelligibility and naturalness of speech. However, it needs more storage and is difficult to realize on small size equipment. In this paper, based on research on the A type compress and applying the adaptive quantization and adaptive predication, the coding compress is achieved to reduce the storage about one half, and the augment of complexity of speech synthesis is not large. The intelligibility and naturalness of the synthesized speech is preserved. And this enlarges the application of speech synthesis.