小波包融合CWT的运动想象脑电信号识别算法

Wavelet Packet Fusion CWT Algorithm for Motor Imagery EEG Signal Recognition

针对传统时频分析方法处理运动想象脑电信号时存在时频分辨率差和分类正确率低的问题,提出一种基于小波包变换(WPT)并优化连续小波变换(CWT)的特征提取算法。对原始信号进行4层小波包分解,提取与运动想象执行时事件相关同步/去同步(ERD/ERS)显著频段的小波能量特征,重构该频段数据作为CWT的输入数据,利用CWT得到最优时段下的时频特征,通过支持向量机(SVM)对提取的特征集进行分类,对比验证单特征与融合特征在整个时段和最优时段的平均识别率。在BCI Competition II的Data III数据集中,该方法在最优时段的平均识别率为89.04%,最高识别率达到91.68%,验证了小波包融合CWT特征提取算法的有效性。

Aiming at the problems of poor time-frequency resolution and low classification correctness in traditional time-frequency analysis methods for processing motor imagery EEG signals,a feature extraction algorithm based on the wavelet packet transform(WPT)and optimizing the continuous wavelet transform(CWT)is proposed.A 4-layer wavelet packet decomposition is performed on the original signal,and significant frequency bands of the event-related synchronization/de-synchronization(ERD/ERS)associated with the execution of motor imagery are extracted,and then the data in this frequency band are reconstructed as the input data for CWT.CWT is used to get the time-frequency features under the optimal time period,and finally the extracted feature set is classified by Support Vector Machine(SVM),comparing and verifying the average recognition rate of the single feature and the fusion feature in the whole time period and the optimal time period.In the Data III dataset of BCI Competition II,the average recognition rate of this method in the optimal time period is 89.04%,and the highest recognition rate reaches 91.68%,which verifies the effectiveness of the feature extraction algorithm of wavelet packet fusion CWT.