论文部分内容阅读
在非植入式脑-机接口(BCI)研究中,独立分量分析(ICA)一直被认为是具有很大应用前景的脑电(EEG)预处理和特征增强方法,但到目前为止,有关在线ICA-BCI系统的研究与实现的报道还不多见。本文对基于ICA的运动想象BCI(MIBCI)系统进行研究,结合ICA无监督学习特点和运动相关去同步化(ERD)现象,构建了一种简单实用的ICA空域滤波器设计方法和三类运动想象判别准则。为了验证所提算法的在线处理性能,本文基于Neuro Scan脑电采集系统和VC++软件平台,完整地实现了在线ICA-MIBCI实验系统。4名受试者参加了系统测试实验,其中两名受试者参加了在线模式的实验。离线和在线实验的三分类运动想象识别结果分别达到了89.78%和89.89%。实验结果表明,本文所提算法分类正确率高,时间开销小,具备跨平台移植的潜力。
In non-implantable brain-computer interface (BCI) studies, independent component analysis (ICA) has long been considered as a promising method of EEG preprocessing and feature enhancement. However, up to now, The research and implementation of ICA-BCI system is still rare. In this paper, ICA-based motion imaging BCI (MIBCI) system is studied. Combining the unsupervised learning features of ICA and motion-related desynchronization (ERD), a simple and practical ICA spatial filter design method and three types of motion imaging Criteria for discrimination. In order to verify the online processing performance of the proposed algorithm, based on the Neuro Scan EEG acquisition system and VC ++ software platform, this paper realizes the online ICA-MIBCI experiment system completely. Four subjects participated in a system testing experiment in which two subjects participated in an experiment in an online mode. The results of three-class motion imaging recognition of offline and online experiments reached 89.78% and 89.89% respectively. Experimental results show that the proposed algorithm has the advantages of high classification accuracy, small time overhead and potential for cross-platform migration.