基于EMD技术的语音特征提取方法

利用软件MATLAB仿真了语音信号的经验模态分解（EMD）特性，并与相关文献报导的其他分解方法进行了比较，结果表明：EMD法能达到更好的说话人识别效果。当不同的人发相同语音时，其频谱特性是不同的，把语音信号进行EMD分解后的IMF做频谱变换，便能得到一个特征向量，于是根据特征向量的不同而达到说话人识别的目的。以上实验结果有助于开辟说话人识别的新途径。

The Method of Speech Feature Extraction Based on Empirical Mode Decomposition

MATLAB is used to calculate the characteristic of speech signal using empirical mode decomposition(EMD),while compared with other methods,we find that:EMD has a better effect in speaker identification.When different people say the same words,their speech signal has the different characteristic of frequency,so we obtain an eigenvector by transforming the IMF of speech signal,and obtain a new method of speaker identification using EMD.The result is helpful for finding new method of speaker identification.