基于脑机接口竞赛脑电数据集的运动想象识别影响因素分析

针对两个常用的脑机接口(brain computer interfacing,BCI)竞赛运动想象任务脑电数据集——2003Ⅲ和2005Ⅲa,分类识别率差异较大的问题,结合运动想象脑电信号的生理基础,从信号噪声、受试者反应程度和C3、C4通道信号能量差异三个方面对两个数据集的数据进行了分析比较。结果表明:与2005Ⅲa数据集相比,2003Ⅲ数据集信噪比较高,受试者对运动想象任务的反应更明显,并且C3C4通道的信号能量差异趋势更标准,解释了对2003Ⅲ数据集的识别率普遍高于2005Ⅲa数据集这一现象,说明这三个因素是影响识别率的主要因素。研究为运动想象脑电数据有效性分析以及提高基于脑电的运动想象识别率提供了新的思路。

Influencing Factors Analysis in Motor Imagery Recognition Based on BCI EEG Database

Based on two commonly used brain computer Interfacing (BCI) competition motor imagery databases, 2003 III and 2005 IIIa, aiming at the big difference in classification accuracy between the two databases, and combining with the physiological basis of motor imagery electroencephalogram (EEG) signals, the data in two databases is analyzed from three aspects: signal noise, subject response degree and signal energy difference between channel C3 and channel C4. The analysis results show that, the 2003 III database comparing with the 2005 IIIa database, has a higher signal-to-noise ratio, the subjects respond more significantly to the motor imagery task, and the signal energy difference trend between channel C3 and channel C4 is more standard. The study explains that the recognition rate of the 2003 III database is higher than that of the 2005 IIIa database, indicates that these three factors are the main factors influencing the recognition rate, and provides a new idea for analyzing the validity of EEG data and improving the recognition rate of motor imagery EEG.