【摘要】 提出了一种新的说话人识别中特征参数的提取方法.在分别使用傅立叶分析和小波分析得到两组特征参数之后,进一步利用Fisher准则进行参数选取,构造了一种新的混合特征参数.在不增加训练和识别时计算量的同时,结合了傅立叶分析和小波分析两者的优点,具有更好的分类能力.实验结果显示,这种新的混合参数有效地提高了说话人的识别率,能更好地表征说话人的特征.更多还原

【Abstract】 A new feature extraction method for speaker identification is proposed. Fourier analysis and wavelet analysis are used to obtain two kinds of feature parameters, then a new mixed feature is constructed by selecting the parameters with Fisher principle. Combining the advantages of both Fourier analysis and wavelet analysis, the method exhibits better discriminative ability without extra calculation in training and identifying. Experiment result shows the new mixed feature efficiently improves the speaker identification accuracy with better feature characterization.更多还原