Voice conversion , as a important part of speech signal processing , is a method which transforms the source speech to converted speech with character information of target speaker , and in voice conversion , the semantic content and environment information of source speaker are remained . voice conversion refers to signal processing , artificial intelligence , acoustics and other fields . it is a typical creature of subjects crossing and has closely relationship with speech recognition , coding , and synthesis , as well as there is resemblance between voice conversion and speaker verification / recognition 研究表明,在线性多变量回归的改进实验中,由于引入了时间概念,使男声-女声的转换平均时间修正度约为21 . 1 % ;在多项式回归、线性多变量回归和支持向量回归的对比实验中,多项式回归在男声到男声的转换和线性多变量回归在女声到女声的转换中,出现了频谱距离增大的现象,而支持向量回归方法使所有的转换语音都具有更好的目标倾向性。