基于共空间模式滤波和优化复杂网络的多通道注意力检测方法

An attention detection method using multi-channel EEG signals based on common spatial pattern and optimized complex network

在异步脑机接口系统中,检测注意力水平可以作为识别控制和非控制状态的一个有效指标。但是现有算法的注意力检测精度较低,难以满足异步脑机接口的高可靠性要求,为此,本文提出了一种基于共空间模式滤波和优化复杂网络的多通道注意力检测方法,通过共空间模式滤波算法进行空间降维,构建优化参数的复杂网络结构并提取图论特征,用于分类专注和放松状态下的脑电信号。16名受试者的交叉验证结果表明,相比于注意力检测中常用的优化复杂网络方法和α+β+δ+θ+R方法,本文所提方法在数据长度为0.5 s和1 s时都取得了更高的分类准确率。

In an asynchronous brain-computer interface system,the detection of the attentional level provides an effective metric for recognizing the control and non-control states.However,the current algorithms used for attention detection delivered relatively low accuracies,which cannot meet the requirement of asynchronous system.To improve the detection accuracy of the attentional status,an attention detection method using multi-channel electroencephalogram signals based on common spatial pattern and optimized complex network is proposed.The multi-channel signals were spatially filtered using the common spatial pattern algorithm.The spatially filtered data were then processed using the optimized complex network algorithm to construct the complexed network structure.The features of graph theory were extracted for classifying the brainwave patterns of the concentration and relax states.The cross-validation results of 16 subjects show that the proposed method has achieved higher classification accuracies than two commonly used methods,i.e.,optimized complex network andα+β+δ+θ+R,with data length of both 0.5 s and 1 s.