基于CSP的模拟阅读脑-机接口信号分类

EEG classification for CSP-based imitating reading BCI

提出了一种基于模拟阅读事件相关电位诱发模式下脑电信号的特征提取及模式分类方法.对采集到的32通道非靶刺激和靶刺激信号进行低通滤波、下采样等处理,根据脑电信号的空间分布特征,选取若干通道的数据,并使用共空间模式算法进行信号的特征提取,最后使用支持向量机(SVM)对信号进行分类.实验结果表明:用位于大脑皮层后半部分通道的数据进行特征提取和分类能较好地识别出靶刺激信号,五名受试者可以达到的最大平均分类正确率分别为90.60%,83.30%,83.98%,72.61%和93.54%.

A feature extraction and pattern classification were given for the signals of imitating reading brain-computer interface based on common spatial pattern and support vector machine. The non-target stimuli signal and target stimuli signal were preprocessed by a low-pass digital filter and down-sam- pled. According to the distribution of the EEG data, the data of some channels for the feature extrac- tion were selected based on common spatial pattern （CSP） training the support vector machine （SVM）. Finally, the testing dataset was classified by using the model. The experimental results show that the target stimuli signal can identified by use of the data located on the back brain for feature ex- traction and classification. The maximum mean classification accuracy of 5 subjects can achieve is up to 90.60% ,83.30% ,83.98% ,72.61% and 93.54%, respectively.