Abstract: Empirical Mode Decomposition（EMD） is a method of signal analysis for processing nonlinear and non-stationary signal. EMD can decompose out different time scale of local feature from the original signal. And then the Intrinsic Mode Function（IMF） of those local characteristic signals is got. A new feature extraction and selection method of underwater acoustic signals based on energy spectrum of Intrinsic Mode Function（IMF’s） is presented, where these intrinsic mode components are decomposed via empirical mode decomposition from original signals and transformed into energy spectrum feature vectors, and thus the different signals’ energy spectrum features of sub-band frequency can be inspected. Support Vector Machine（SVM） classifier is used for classification experiments. The results show that the correct identification ratio of experiments based on IMF’s energy spectrum is above 88%, which is superior to feature extraction of wavelet energy spectrum.

Key words: Empirical Mode Decomposition（EMD）, Intrinsic Mode Function（IMF）, Intrinsic Mode Function（IMF） energy spectrum, feature extraction, Support Vector Machine（SVM） classifier

摘要： 经验模态分解（EMD）是用来处理非平稳时变信号的一种信号分析方法，该方法对所分析信号的局部特征信号进行不同时间尺度的分解，从而得到这些局部特征信号的各阶本征模函数（IMF）。提出了一种基于IMF能量谱的水声信号特征提取与选择方法，通过对水声信号进行经验模态分解，提取信号的本征模式分量并转换为能量谱特征向量，从而观测不同信号子频带能量谱的特征变化。分类实验采用支持向量机（SVM）分类器进行。实验结果表明，相对于小波能量谱特征提取法而言，利用IMF能量谱作为特征向量的分类实验具有更佳的分类效果，平均正确率达88%以上。

关键词: 经验模态分解, 本征模函数, 本征模函数能量谱, 特征提取, 支持向量机（SVM）分类器