快速序列视觉呈现任务下的脑电分类算法

提出了一个在快速序列视觉呈现任务下的脑电信号分类算法.将图片序列快速呈现给受试者并将同步采集脑电信号，将脑电信号截取分段作为样本集.通过约束有监督降维后样本与样本中心差值的趋近方向，使用训练集脑电数据训练得到映射矩阵；通过特征提取函数将训练集和测试集的脑电数据样本变换为特征矢量，使用支持向量机对样本进行分类.实验结果表明，算法对24名受试者的脑电信号分类的平均正确率为91.5%，平均AUC达到了0.95，证明脑电分类算法具有良好的分类性能，可以在快速序列视觉呈现任务中准确地识别目标图片.

EEG Classification Algorithm for Rapid Serial Visual Presentation Task

In this project, we proposed a classification algorithm of electroencephalogram (EEG) signals in order to fulfill the Rapid Serial Visual Presentation (RSVP) task. Firstly, the EEG signals of the subjects were recorded when they received the image sequences and then segmented to creat a sample set. Secondly, by confining the difference between the sample and the sample center after supervised dimensionality reduction, the mapping matrix was obtained after training EEG data from the training set. EEG samples of training set and test set were transformed into feature vectors by using feature extracting function, and support vector machine (SVM) was used to classify the EEG samples. The experiment results showed that the average classification accuracy rate of EEG of 24 subjects was 91.5% and the average AUC was 0.95, which indicates that the EEG classification algorithm has good classification performance and can accurately detect target images in the Rapid Serial Visual Presentation tasks.