基于改进HHT和样本熵的脑电信号特征提取

摘要

After processing the motor imagery electroencephalogram (EEG) signal by empirical mode decomposition (EMD) ,in‐trinsic mode functions (IMF) are artificially selected ,which results in mixing the reconstructed signal with noise and lossing use‐ful information .Therefore the feature extraction method of combining improved Hilbert‐Huang transform (HHT) and sample entropy was proposed .After the original EEG signal was processed using EMD ,each IMF and the original signal were used to calculate the correlation coefficient and the number of IMF’s instantaneous frequency belonging to the μ/βrhythm band .The average energy of the effective IMF was extracted and the sample entropy of EEG signal was calculated to constitute the feature vectors .The extracted features were classified by using support vector machine (SVM ) classifier .And the algorithm was veri‐fied on the smart wheelchair platform .The results show that the wheelchair system based on improved HHT and sample entropy has higher correct recognition rate and better stability .%针对运动想象脑电信号在经验模态分解（EMD）后人为选取固有模态函数（IMF）导致重构信号混入噪声且丢失有用信息的问题，提出一种改进希尔伯特‐黄变换（HHT）和样本熵结合的特征提取方法。在原始脑电信号经过EMD后，计算各个IM F与原始信号的相关系数以及IM F中瞬时频率在μ／β节律频带内的个数，提取有效IM F的能量均值，联合计算脑电信号的样本熵构成特征向量，采用支持向量机（SVM ）分类器对提取的特征进行分类，在智能轮椅平台上对算法进行验证。验证结果表明，采用改进 HHT结合样本熵的智能轮椅系统有较高正确识别率，稳定性更好。