Abstract: A new feature extraction and selection method based on the energy entropy of intrinsic mode functions（IMF’s） is presented.Three types of noise signals radiated from the targets are decomposed into their respective IMF’s using the Empirical Mode Decomposition procedure，and the energy of the same IMF of three types of signals are different.The energy entropies of the IMF’s are calculated by Eq（5） or Eq（6）.K-neighbor classifier is used for classification experiments for three types of signals.The results show that the correct identification ratio of experiments based on esq.（6） is above 80%.

Key words: Empirical Mode Decomposition（EMD）, Intrinsic Mode Function（IMF）, feature extraction, K-neighbor classifier

摘要： 提出了一种基于固有模态函数（IMF）能量熵的特征提取与选择方法。对三类信号进行了经验模态分解（EMD），得到IMF。对于不同类别的信号，同阶的IMF能量有明显的不同。选择IMF能量作为特征向量，并选判别熵作为分类判据，同时给出了两种能量熵的计算公式。采用K－近邻分类器对三类信号进行了分类试验，试验结果表明，基于最佳特征向量选择的分类试验的平均正确识别率达80%以上。

关键词: 经验模态分解, 固有模态函数, 特征提取, K－近邻分类