基于支持向量多分类机的多类复杂手操作EEG信号模式识别

针对用于服务机器人的脑机接口系统中脑电信号模式识别精度不高,不能满足机器人多任务要求的问题,提出一种基于C-支持向量多分类机的多类复杂手操作EEG信号模式识别方法,并将其应用到复杂手操作的EEG信号模式识别试验中,实现一个4类复杂手操作的模式识别,实验结果表明,与之前用BP神经网络进行识别相比,识别率由85%提高到了90%。

Since the low EEG pattern recognition precision of the service robot s brain-computer interface system can not meet the robot s multi-tasks,in this paper,a EEG pattern recognition method for multi-complicated hand activities based on C-support vector classifiers was proposed and putted into the EEG pattern recognition for multi-complicated hand activities experiment which include four classes.The result show the recognition rate rise to 90% comparad to the 85% by using the BP neural network.

机器人学国家重点实验室开放课题(RLO200801); 江苏省自然科学基金(BK2007059)