Abstract

Info

Author(s):: WU Yelan; ZHANG Yue; CAO Pugang; LIAN Xiaoqin; YU Chongchong; School of Artificial Intelligence, Beijing Technology and Business University, Beijing 100048, China

Keywords:: Keywords: motor imagery electroencephalogram feature extraction filter bank common spatial pattern immune particle swarm optimization

Abstract:: Abstract: Objective To propose a method for spatial feature extraction based on rhythm adaption for addressing the problem of poor recognition of multi-class motor imagery electroencephalogram (MI-EEG) caused by the individual differences in MI-EEG and the dependence of feature quality on frequency band selection. Methods The spatial features under different frequency bands were extracted with filter bank common spatial pattern (FBCSP). The immune particle swarm optimization algorithm was used to optimize the frequency band and spatial feature extraction parameters in feature extraction for realizing the adaptive adjustment of the rhythm and spatial parameters and obtaining the FBCSP spatial features under the optimal rhythm, thereby improving the recognition accuracy of multi-class MI-EEG. Results The proposed method had an average accuracy of 85.49% on BCI-IV Dataset 2a and BCI-Ⅲ Dataset 3a, which was 10.84% higher than the original FBCSP method. Conclusion The proposed method is advantageous in EEG feature extraction and can effectively improve classification accuracy, providing a new solution for MI-EEG classification.