Abstract: As the amateur singer’s timbre is not converted in the process of voice imitation from amateur singer to the professional singer, this paper presents a singing voice Morphing algorithm based on Gaussian Mixture Model(GMM). The spectrum parameters are extracted from the source and the target singing voices are modeled by GMM respectively, and the singing’s timbre is converted by mixing their model parameters. Experiments demonstrate that the accuracy of ABX test can be up to 100% in the case of k=0 or 1, and the accuracy still be higher than 62.5% in the case of 0

Key words: STRAIGHT algorithm, timbre modification, Gaussian Mixture Model(GMM), singing voice Morphing

摘要： 针对业余歌手模仿专业歌手唱歌过程中音色不变的问题，提出一种基于高斯混合模型(GMM)的中文歌曲Morphing算法，采用GMM对语音频谱建模，并通过混合业余歌手和专业歌手的语音频谱，实现歌曲的音色转换。结果显示，混合比例因子k=0或1时，ABX测试正确率均为100%，0

关键词: STRAIGHT算法, 音色修改, 高斯混合模型, 歌曲Morphing

CLC Number: 

• TP391