论文部分内容阅读
首先建立了由维吾尔语中的单音素、双音素所构成的小规模语音语料库,设计了相应的拼接单元挑选算法,利用参数调整算法对拼接单元语音信号的时长、基频和短时能量等特征参数进行调整,并利用时域平滑算法对拼接点处的语音参数进行调整,从而进一步提高了合成语音的自然度。用C Sharp编程语言实现了上述算法,试验结果表明研究思路和技术方案的可行性。该系统具有语料库小,合成语音的可懂度和自然度较高等优势。
Firstly, a small-scale phonetic corpus composed of mono-phonemes and diphones in Uyghur language is established, and the corresponding splicing unit selection algorithm is designed. Parameters such as duration, fundamental frequency and short-time energy of the spliced unit speech signal are adjusted by parameter adjustment algorithm Adjust the parameters, and use the time-domain smoothing algorithm to adjust the speech parameters at the splicing point so as to further improve the naturalness of the synthesized speech. The above algorithm is implemented with C Sharp programming language, and the experimental results show the feasibility of the research ideas and technical solutions. The system has the advantages of small corpus, intelligibility of synthesized speech and high natural degree.