基于ERD/ERS现象的运动想象脑电信号分类研究

Research on classification method based on ERD/ERS phenomenon for the motor imagery eeg

运动想象脑电信号的分类正确率直接影响到BCI系统的应用价值.为了提高分类正确率,通过对脑电信号的特征提取和分类识别方法的研究,根据运动想象过程中的ERD/ERS现象,提出利用独立分量分析法对采集到的多导联运动想象脑电信号进行盲源分离,获取与想象任务相关的ICA滤波器,将其作用到预处理后的观测信号上,得到与想象任务相关的脑电成分,再以支持向量机为分类器来实现对三类运动想象任务的脑电信号进行分类识别的方法,并对三位受试者的脑电数据进行了分类实验,实验结果验证了该方法的有效性.

The correct rate of classification of motor imagery EEG signals has a direct impact on the application of BCI system.In order to improve the correct rate,a study on feature extraction and classification of EEG signals has been made.According to the ERD/ERS phenomenon,the method of independent component analysis has been used to make blind source separation and get ICA filters which are used on the observed multi-channel signals to find components associated with motor imagery task.Then,support vector machine has been used for data classification and recognition of three-class motor imagery EEG.The classification experiments have been carried out on three subjects.Experimental results show that this method is simple and effective.