三类运动想象脑电信号的离线分析研究

The research of off-line analysis of three-class motor imagery EEG

为解决运动想象脑电信号(EEG)的多分类问题,本文提出了一种基于粒子群优化支持向量机(PSO-SVM)的EEG分类方法,采用NEUROSCAN平台设计实验自测数据,对想象左手握握力器,右手握握力器,右脚踩油门三类运动想象任务进行了分类识别研究。采用FFT和IFFT对信号进行预处理,采用离散小波分析(DWT)提取能量值,并结合小波系数作为组合特征,分类效果明显好于BP和自组织神经网络(SOM)分类器。

To solve the problem of motor imagery EEG multi-classification,we designed an experiment to get EEG data through NEUROSCAN,and proposed a classification method based on Particle Swarm Optimization-Support Vector Machine to recognize the three kinds of motor imagery tasks,including left hand grip-grasped,right hand grip-grasped and right foot accelerator-stepped.The signal pre-processing was conducted via FFT and IFFT.The discrete wavelet transform was used to extract the feature,The results show that the proposed classifier is obviously better than BP classifier and self-organizing map。