一种基于特征差异度和SVM投票机制的数字音乐语音情感识别算法

针对数字音乐语音情感识别问题,提出一种基于特征差异度和SVM投票机制进行识别的方法.该方法不仅降低了特征向量的维度,而且保留了足够的能够描述数字音乐语音不同情感之间差异的特征.同时,该方法利用多个二分SVM分类器进行投票,减少了每个分类器的权重,从而降低了误差.实验结果表明,该方法能够有效地提高识别准确率.

An emotion recognition algorithm for digital music speech based on difference between characteristics and SVM voting mechanism

To solve emotion recognition problem for digital music speech, this paper proposes an emotion recognition method which is based on difference between characteristics and SVM voting mechanism. The method not only reduces the dimension of feature vectors, but retains the characteristics between different emotions of digital music speech. Meanwhile, using multiple two-class SVM classifier to vote for digital music speech, the method decreases the weight of each classifiers and reduces the error. The experimental results show that this method can effectively improve the recognition accuracy.