Abstract: In order to effectively classify and detect electroencephalogram signals in stroke patients with midbrain infarction and cerebral hemorrhage, the article proposes an automatic classification prediction method. The method bases on the feature fusion of wavelet packet energy and approximate entropy feature. Firstly, the input signal is decomposed for getting the energy of each frequency bands, then the energy dimension is reduced. After that the energy after dimension reduction is fused with the signal’s approximate entropy to get the final features. Lastly, support vector machine algorithm is used to train the prediction model for classifying the stroke input data. The research results show that this method has good ability of EEG feature signal classification. Moreover, extracting alpha band EEG signal alone as input signal, the experimental results show that especially the high rate can be obtained. The prediction accuracy of cerebral infarction and cerebral hemorrhage can reach 98.36% in average. It plays a good auxiliary decision-making role in the clinical prediction and detection of stroke diseases.

Key words: electroencephalogram, stroke, alpha band, wavelet packet energy, approximate entropy

摘要： 为了实现脑卒中患者中脑梗死、脑出血两类疾病脑电信号的高效分类与检测，提出了一种基于小波包能量与近似熵特征提取结合的脑电自动分类预测方法。将输入的脑卒中病人的脑电信号进行小波包分解，提取各个频段的能量并降维，而后与近似熵融合作为特征向量，并用支持向量机算法对其进行分类。研究结果表明该方法有较强的脑电特征分类识别能力，进一步单独提取原始脑电信号α波段的信号进行分类，得到了更优的分类效果，脑卒中脑电信号的分类准确率可以达到98.36%。这对临床上实现脑卒中疾病的智能预测具有辅助决策作用。

关键词: 脑电图, 脑卒中, &alpha, 波段, 小波包能量, 近似熵