文献类型：期刊文献

英文题名：Coefficient-of-variation-based channel selection with a new testing framework for MI-based BCI

作者：Xiao, Ruocheng[1];Huang, Yitao[1];Xu, Ren[2];Wang, Bei[1];Wang, Xingyu[1];Jin, Jing[1]

机构：[1]East China Univ Sci & Technol, Key Lab Smart Mfg Energy Chem Proc, Minist Educ, Shanghai, Peoples R China;[2]Guger Technol OG, Graz, Austria

年份：2022

卷号：16

期号：4

起止页码：791

外文期刊名：COGNITIVE NEURODYNAMICS

收录：;WOS:【SCI-EXPANDED(收录号:WOS:000723511500001)】；

基金：The authors are grateful for the support of the National Key Research and Development Program (2017YFB13003002), the Grant National Natural Science Foundation of China (61573142, 61773164, and 91420302), and the program of Introducing Talents of Discipline to Universities (the 111 Project) under Grant B17017.

语种：英文

外文关键词：Brain-computer interface (BCI); Electroencephalogram (EEG); Motor imagery (MI); Channel selection; Coefficient of variation (C; V; )

摘要：In the motor-imagery (MI) based brain computer interface (BCI), multi-channel electroencephalogram (EEG) is often used to ensure the complete capture of physiological phenomena. With the redundant information and noise, EEG signals cannot be easily converted into separable features through feature extraction algorithms. Channel selection algorithms are proposed to address the issue, in which the filtering technique is widely used with the advantages of low computational cost and strong practicability. In this study, we proposed several improved methods for filtering channel selection algorithm. Specifically, based on the coefficient of variation and inter-class distance, a novel channel classification method was designed, which divided channels into different categories based on their contribution to feature extraction process. Then a filtering channel selection algorithm was proposed according to the previous classification method. Moreover, a new testing framework for filtering channel selection algorithms was proposed, which can better reflect the generalization ability of the algorithm. Experimental results indicated that the proposed channel classification method is effective, and the proposed testing framework is better than the original one. Meanwhile, the proposed channel selection algorithm achieved the accuracy of 87.7% and 81.7% in two BCI competition datasets, respectively, which was superior to competing algorithms.