基于参数模型和FastICA算法的P300特征实时提取  

Real-Time P300 Feature Extraction Based on Parametric Model and FastICA Algorithm

作　　者：李窦哲[1] 乔晓艳[1] 董有尔[1] 

机构地区：[1]山西大学物理电子工程学院,山西太原030006

出　　处：《测试技术学报》2009年第6期505-509,共5页Journal of Test and Measurement Technology

基　　金：山西省自然科学基金资助项目(2007011041);山西省高校科技开发资助项目(20081001)

摘　　要：利用AR模型结合快速独立成份分析(F astICA)对视觉诱发电位P 300特征进行实时提取.采用图片轮换作为诱发刺激获取EEG信号,利用主成份分析对脑电信号降维处理,然后进行独立成份分析,计算各独立成份与垂直眼电导联的相关性,自动去除眼电干扰.利用AR模型对自发脑电建模,并除去EEG中的自发脑电,通过少次相干平均提取P 300.实验结果表明:该方法可以不依赖任何先验信息,在不降低信噪比的情况下实时提取P 300特征,避免了长时间视觉刺激引起视觉疲劳而产生的误差,提高了脑机接口系统的识别效率,为建立在线BC I系统奠定了基础.A method based on AR model and FastICA algorithm for P300 feature extraction is presented. In the study, the visual evoked potential is obtained via the alternate pictures. Then, principal component analysis （PCA） is used for reducing the dimensions of EEG signal, independent component analysis （ICA） for removing EOG artifact. And the AR model is constructed for filtrating the spontaneous EEG. Finally, a coherence average is used to extract P300 in real-time. The results show that this method can perform effectively to extract P300 feature without depending on any prior information and avoid the subject＇s visual fatigue caused by long time visual evoke. It can improve the recognition efficiency of BCI system and be applied to the online form.

关 键 词：特征提取 P300 独立成份分析 参数模型 

分 类 号：TP274.2[自动化与计算机技术—检测技术与自动化装置]