Cross-Language Speech Emotion Recognition via Multiple Kernel Learning

Abstract

Due to the difference of the speaker's language, speech emotion recognition tasks often face the situation that training data are not fully representative of test data. Therefore, the space extended by a kernel function. might not sufficient to describe different properties of data and thus produce a satisfactory decision function. In this wok, we apply multiple kernel learning to recognize the speech emotion of cross-language. Compared to SVM, multiple kernel learning can achieve better performance in cross-language speech emotion recognition tasks.