基于样本熵的运动想象分类研究

Classification of Motor Imagery Based on Sample Entropy

摘要: 提出了基于脑电的样本熵特征进行运动想象分类的思想，分析了左右手运动想象时感觉运动皮层的脑电信号样本熵及其动态变化规律．结果表明，样本熵能够较好地反映左右手运动想象时脑电特征的变化，具有明确的生理意义．在此基础上，利用Fisher线性分类器对基于样本熵的左右手运动想象进行了动态分类，得到的平均最大分类正确率达到87.8%．最后，提出了一种样本熵的快速算法，其计算量小、速度快，满足BCI实时系统要求．

Abstract: The classification method of motor imagery based on sample entropy(SampEn) of electroencephalogram(EEG) is proposed.The SampEn of EEG in primary sensorimotor area and its dynamic properties during left-right hand motor imagination are analyzed.Experiment results show that SampEn can reflect the EEG pattern changes of left-right hand motor imageries and have clear physiological explanation.Fisher LDA(Linear Discriminant Analysis) is used to dynamically classify the left-right hand movement imageries based on SampEn features,and an average maximum classification accuracy of 87.8% is obtained.Finally,a fast algorithm of SampEn with minimum computation cost and high speed is introduced,which can meet the requirements of real-time brain-computer interface(BCI) system.