建立了普通话语音性别数据库,提出联合梅尔频率频谱系数(Mel-frequency Cepstrum Coefficients,MFCC)的特征提取方法和支持向量机(Support Vector Machine,SVM)的分类方法进行说话人性别识别,并与其它分类方法进行比较,实验结果表明该方法的说话人性别识别准确率达到98.7%,明显优于其它分类器。
A Chinese speech (mandarin) database was established for speakers gender recognition. A combination method is proposed for gender recognition of speakers based on support vector machine and Mel-frequency cepstrum coefficients (MFCC) for classification and feature extraction respectively. The comparative result shows that the accuracy of SVM is 98.7%, which is better than other methods.