一种用于运动想象脑电信号的混合特征选择算法  

A hybrid feature selection algorithm for motor imagery electroencephalogram signals

作　　者：刘紫恒 周建华[1] LIU Zi-heng;ZHOU Jian-hua(Faculty of Information Engineering and Automation,Kunming University of Science and Technology,Kunming 650031,China)

机构地区：[1]昆明理工大学信息工程与自动化学院,昆明650031

出　　处：《兰州大学学报（自然科学版）》2024年第2期167-172,共6页Journal of Lanzhou University(Natural Sciences)

基　　金：国家自然科学基金项目(82172058)。

摘　　要：针对过滤式特征选择算法精度不高和包裹式特征选择算法训练时间长的缺点,提出一种融合信息增益(IG)和自适应遗传算法(AGA)的混合特征选择算法.用滤波器组公共空间模式提取运动想象脑电信号特征,计算每个特征的IG并排序,根据排序用阈值法剔除部分无用特征,用AGA在剩余特征中搜索出最优特征子集.用2个公共数据集验证所提出算法的有效性,取得81.24%±15.04%的平均分类准确率,平均用时3.68 s.所提出算法的分类准确率大于过滤式算法,训练时长短于包裹式算法.In view of the problems of low accuracy with filter feature selection algorithm and long training time of wrapper feature selection algorithm,a hybrid feature selection algorithm based on mixed information gain and adaptive genetic algorithm was proposed.The filter bank common spatial pattern was used to extract the features of motor imagery electroencephalogram signals,and the information gain of each feature calculated and sorted.According to the sorting,the threshold value method was used to eliminate some useless features.The optimal feature subset sought out from the remaining features was classified via the adaptive genetic algorithm.Two common data sets were used to verify the validity of the algorithm,and the average classification accuracy was 81.24%±15.04%and the average time was 3.68 s.The experimental results showed that the classification accuracy of the proposed algorithm was higher than that of the filter algorithm,and the training time was shorter than that of the wrapper algorithm.

关 键 词：运动想象 特征选择 信息增益 自适应遗传算法 

分 类 号：TN911.7[电子电信—通信与信息系统]