基于特征语音的说话人自适应算法研究

Speaker adaptation algorithm based on eigen voice

介绍了说话人自适应技术中的特征语音(Eigenvoice,EV)方法。用最大后验概率特征分解(Maximum a Posteriori Eigen-decomposition,MAPED)法来计算线性组合系数,代替了传统方法中的最大似然特征分解(Maximum Likelihood Eigen-decomposition,MLED)的方法。实验对这两种方法的性能进行了比较。结果证明使用MAPED这种方法比用MLED的方法错误识别率有一定的降低,增强了系统的鲁棒性。

This paper discusses on an algorithm of eigen voice （EV） in the field of speaker adaptation. It improves the standard of EV speaker adaptation method by using maximum a posteriori eigen - decomposition （MAPED）. The experiments show that MAPED is able to achieve better performance than maximum likelihood eigen- decomposition （MLED） with few adaptation data, and is well suited for the robust speech recognition.