音色感知不变的参数调节与分析

Parameters adjustment and analysis of speech timbre's perceptual invariance

讨论了在语音音色修饰技术中保持原始音色主观感知的2种参数的调节范围。分别录制短句、长句作为原始语料,利用STRAIGHT模型对原始语料的基频F0和音节时长这2个参数进行调节。对调节后的合成语音进行主观实验,确定在发音人音色不发生较大改变的原则下所获得的参数调节范围。最后利用MOS评分对该参数范围进行了符合自然度听感要求的优化压缩。

In this paper, the adjustment range of two parameters which are related to the subjective perception of timbre in speech modification technology are discussed. Phrases and sentences are separately recorded as the original corpus, then their F0 and syllable duration are adjusted by using STRAIGHT. After the subjective experiment of the synthesized speech, the adjustment range of the parameters is determined with the principle that the timbre perception does not change greatly. Finally, the MOS score is used to optimize the adjustment range of the parameters.