核主分量分析(Kernel principal component analysis ,KPCA)是一种利用核方法将主分量分析(Principal component analysis,PCA)推广后的学习方法,KPCA方法能够使得输入空间线性不可分的样本在特征空间有更好的可分性。典型相关分析(Canonical correlation analysis,CCA)是分析两组随机变量之间的相关性的一种统计方法。本文提出将KPCA方法用于语音情感识别中,并采用KPCA和CCA结合的方法用于情感识别。与传统的PCA方法进行了对比,研究结果表明基于KPCA及KPCA+CCA的情感识别有较好的效果。
Kernel principal component analysis (KPCA) is a learning method that uses the ker- nel method on principal component analysis (PCA), and the KPCA method can make the in- separable samples separable in the feature space. Canonical correlation analysis (CCA) is a sta- tistical analysis of the correlation between the two sets of random variables. In this paper, the KPCA method is proposed for speech emotion recognition, and the KPCA combined with CCA method is also proposed for emotion recognition. Compared with the traditional PCA method, the results show that the emotion recognition based on KPCA and KPCA + CCA have better performance.