This paper proposes the natural sounding Uyghur speech synthesis method using context dependent syllables for the concatenating units.Context dependent syllables effectively incorporate the sound associated influence of both inside and outside neighboring syllables and can avoid most of the impact of co-articulation.In order to reduce the distortion in concatenation points, a time domain smoothing method is presented for smoothing the synthesized speech.Test indicates that this method can reduce some distortion in the concatenation point and achieve higher naturalness.%提出上下文音节为拼接单元,生成自然语音的一种维吾尔语语音合成方法.上下文音节有效地整合了邻近音节内外的音联现象,可以避免大部分协同发音影响.采用基于时域平滑技术对合成语音进行平滑处理,减少在拼接点处的语音失真和突变,从而提高了合成语音的自然度.
展开▼