脑-机接口信号中靶标的贝叶斯搜索  

Target Searching in EEG Using Bayesian Classifier for Brain-Computer Interface

作　　者：官金安[1] 许珍荣[1] 方浩[1] 

机构地区：[1]中南民族大学电子信息工程学院信号检测与处理实验室,武汉430074

出　　处：《中南民族大学学报（自然科学版）》2009年第4期86-89,共4页Journal of South-Central University for Nationalities：Natural Science Edition

基　　金：国家自然科学基金资助项目(30370393);湖北省自然科学基金资助项目(2007ABA098)

摘　　要：提出了脑控虚拟键盘上按钮的编码以每个按钮中靶标出现的时间次序安排,搜索出靶信号出现的位置就能探测受试者的选择.以基于最小误判准则的贝叶斯分类器,对靶及非靶两类刺激信号进行分类训练,分别在通道和时段选择上进行了特征提取研究,得到识别率最高的5个通道中最佳时段的训练参数.利用训练了的分类器,对受试者每次按钮选择所记录到的EEG信号中可能出现靶刺激的位置进行搜索,来判断受试者所选择的按钮.结果表明:最佳通道P 8获得了72.50%的平均正确识别率,该结果为构建一个实时的脑控拼写装置打下了基础.The button of brain-controlled virtual keyboard is encoded by the time sequence of target on each button.The choice of subjects could be determined by picking out the location of ERP evoked by target.Bayesian classifier based on minimum misjudgment criteria was trained by the target and non-target stimulation signals.After studying the feature extraction of signals based on the channels and time slots,respectively,the training parameters with higher recognition rate were obtained in five channels.Then,the trained classifier was applied on EEG to determine the selected button of user.The results show that the channel of P8 gains the best average correct recognition rate of 72.5 %.The study lays a foundation for building a real-time brain controlled spelling device.

关 键 词：诱发电位 特征提取 靶标搜索 贝叶斯分类器 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]