基于匹配滤波器和度量学习的脑电信号分类  

EEG classification based on a match filter and metric learning

作　　者：刘宏马 王生进[1] LIU Hongma;WANG Shengjin(National Laboratory for Information Science and Technology,State Key Laboratory of Intelligent Technology and Systems,Department of Electronic Engineering,Tsinghua University,Beijing 100084,China)

机构地区：[1]清华大学电子工程系,智能技术与系统国家重点实验室,信息技术国家实验室,北京100084

出　　处：《清华大学学报（自然科学版）》2021年第3期248-253,共6页Journal of Tsinghua University(Science and Technology)

基　　金：国家“八六三”高技术项目(2012AA011004);国家科技支撑计划项目(2013BAK02B04);教育部自主科研计划项目(20141081253)。

摘　　要：脑电信号识别和脑机接口技术是人机交互领域的热点问题。当前脑电信号分类方法模型复杂,难以实际应用。该文提出基于匹配滤波器的脑电信号分类框架:根据脑电信号特点和假设检验建立生成式模型,并基于Gauss噪声假设推导出一个简单的线性判定算子;利用度量学习方法估计主信号分量和最优协方差矩阵,进一步增强分类器的鉴别力。实验结果表明:所推导出的线性判定算子分类精度和计算复杂度都优于其他算法,能够满足实际应用需求。Brain signal analyses and brain-computer interfaces are key topics in human-computer interaction research. Current electroencephalography(EEG) signal classification methods are complicated and difficult to apply in practice. This paper presents a match filter based classification framework using a hypothesis testing model and a match filter which is a linear function of the signal due to the Gaussian noise assumption. A metric learning based method is then used to estimate the principle component and the optimal covariance matrix to further enhance the model discrimination. The results show that this method provides better recognition accuracy with less computational complexity than other algorithms which makes it more practical.

关 键 词：脑机接口 匹配滤波器 度量学习 P300打字机 

分 类 号：TP399[自动化与计算机技术—计算机应用技术]