Abstract: Aiming at the problem that difficulty of feature extraction and low recognition accuracy rate, using Wavelet Packet Entropy （WPE） to feature extraction and Support Vector Machines （SVM） to classification is proposed. First, the power of motor imagery EEG data is calculated, and the scale of wavelet packet is selected by theoretical analysis. Then, the wavelet packet decomposition on power is done, the WPE of power is caculated, and the wavelet packet entropy interpolation of C3, C4 lead is extracted, which compose feature vector. Finally, the feature vector is fed as classifier input into support vector machine to achieve classification. From Graz’s EEG data of international BCI competition 2003, the highest accuracy rate of classification is 97.56%. The feature vectors of this algorithm are low in dimension and small in data size, also have high classification accuracy, which provides a reference method for the task of EEG feature extraction and classification.

Key words: wavelet packet entropy, support vector machine, classification of EEG signal

摘要： 针对运动想象脑电信号特征提取困难，分类正确率低的问题，提出了利用小波熵进行特征提取并采用支持向量机（SVM）来分类的算法。计算运动想象脑电信号的功率，通过理论分析选择小波包尺度，对信号功率进行小波包分解并计算其小波包熵（WPE），提取C3、C4导联的小波包熵插值组成特征向量，将特征向量作为分类器的输入送入支持向量机进行分类。采用国际BCI竞赛2003中的Graz数据进行验证，算法的最高分类正确率达97.56%。算法特征向量维数低、数据量小、分类正确率高，对运动想象脑电信号特征提取及分类的任务可以提供参考方法。

关键词: 小波包熵, 支持向量机, 脑电信号分类