【作者】 蔡莲红； 侯杰； 刘瑞； 徐俊； 张安娜； 王宽； 刘宇翔；

【Author】 Lianhong Cai, Jie Hou, Rui Liu, Jun Xu, Anna Zhang, Kuan Wang, Yuxiang Liu Key Laboratory of Pervasive Computing, Ministry of Education Tsinghua National Laboratory for Information Science and Technology Dept. of Computer Science and Technology, Tsinghua University, Beijing 100084, China

【机构】 普适计算教育部重点实验室； 清华信息科学与技术国家实验室(筹)； 清华大学计算机科学与技术系；

【摘要】 歌唱合成是热门的语音合成应用之一。本文实现了一种基于音高指导的HMM(Hidden Markov Model,隐马尔科夫模型)参数化歌唱合成系统,并提出了若干提升歌唱合成效果的策略。该系统可以根据乐谱音高对每一个音节选择较合适的HMM模型,减小所选取的朗读语音模型和目标歌唱语音模型的基频差异,保证合成参数中基频和频谱的一致性,从而有助于提高合成声音的音质和保持说话人的音色。主观评测实验证明本文的方法有效地提升了歌唱合成的效果。更多还原

【Abstract】 Singing synthesis is one of the hot topics of the application of speech synthesis. This paper introduced a HMM (Hidden Markov Model) parametric singing voice synthesis system with pitch instruction and proposed several strategies that can enhance the effect of singing synthesis. Compared to existing systems, our system has the advantage of being able to select appropriate acoustic models for the synthesis procedure according to the pitch obtained from the reference score, which contributes to reduce the modification of the fundamental frequency and keeps a better consistency of resulting pitch and spectra for better quality. Subjective evaluation proved that our system was effective to enhance the quality of synthesized singing voice.更多还原