Abstract:

In this paper，we use Support Vector Machine with radial basis function to classify the accents of Mandarin Speech which includes the native speakers and non-native speakers from Dai，Lisu and Naxi in Yunnan.In the feature-extraction strategy，we estimate Mel frequency cepstral coefficients，pitch and energy，and calculate their correlative statistical features.Experimental results show that this scheme can achieve an average accuracy of 93%.

Key words: accents recognition, Support Vector Machine（SVM）, Gaussian radial basis function

摘要： 基于母语分别为傣语、傈僳语、纳西语和汉语普通话的发音人所发汉语普通话语句，利用支持向量机进行民族口音识别研究。实验结果表明，采用对每个语句提取12维MFCC参数、9维基频派生参数和9维短时平均能量派生参数等作为特征参数集及使用高斯径向基函数支持向量机的方法，男、女声的口音识别率均超过93%。

关键词: 口音识别, 支持向量机（SVM）, 高斯径向基函数