Abstract: Syllable is the smallest pronunciation unit in Uighur language and the basic synthesis unit in conventional Uighur speech synthesis system. However, the number of syllable is unlimited and speech corpus hardly includes all possible syllables, therefore speech synthesis performances with syllable units are unstable and discontinuous. To solve the instability of synthetic speech, this paper proposes a unit selection algorithm that combines the phoneme and triphone unit. Best prosodic matching unit is selected and discontinuity of synthetic speech is solved by adding the prosodic parameter matching method into unit selection stage. The evaluation results show that the proposed method improves the naturalness of synthetic speech.

Key words: Uighur speech synthesis, phoneme, triphone, prosodic model, unit selection

摘要： 音节是维吾尔语的最小发音单元，所以大部分维吾尔语语音合成系统以音节作为基本的合成单元，但维吾尔语中音节数量很大，语料库很难保证覆盖所有的音节样本，这会导致合成语音不稳定和不连续。为解决合成语音不稳定的情况，提出了结合单音素和三音素两个不同基元的单元挑选算法。通过在单元挑选模块中加入韵律参数相匹配的方法选出最佳韵律匹配的单元并解决了合成语音不连续的情况。实验结果表明，提出的方法有效地解决了合成语音不稳定和不连续的现象，从而提高了合成语音的自然度。

关键词: 维吾尔语音合成, 单音素, 三音素, 韵律参数模型, 单元挑选