Common Spatial Pattern (CSP) is an effective algorithm in constructing optimal spatial filters, which is widely used to discriminate two classes of electroencephalogram (EEG) signal in Motor Imagery (MI) based Brain Computer Interface (BCI). To extend CSP algorithm to three-class motor imagery of left-hand, right-hand, both-feet, in this paper a three-class classification strategy based on Filter Bank Common Spatial Pattern (FBCSP) and voting mechanism is proposed. The strategy reduces a three-class problem to two binary-class problems. Two binary-class classifiers are constructed for the three-class classification, both-hands vs both-feet and left-hand vs right-hand. The result shows an average three-class classification accuracy of 68.6% with BCI competition IV Datasets 2a, which is an encouraging result in motor imagery pattern recognition. And demonstrated that both-hands can be considered as one class in MI based BCI system.
Paper
Full text
Three-class Motor Imagery Classification Based on FBCSP Combined with Voting Mechanism
Semantic Scholar · Computer Science · 2019
Abstract
Common Spatial Pattern (CSP) is an effective algorithm in constructing optimal spatial filters, which is widely used to discriminate two classes of electroencephalogram (EEG) signal in Motor Imagery (MI) based Brain Computer Interface (BCI). To extend CSP algorithm to three-class motor imagery of left-hand, right-hand, both-feet, in this paper a three-class classification strategy based on Filter Bank Common Spatial Pattern (FBCSP) and voting mechanism is proposed. The strategy reduces a three-class problem to two binary-class problems. Two binary-class classifiers are constructed for the three-class classification, both-hands vs both-feet and left-hand vs right-hand. The result shows an average three-class classification accuracy of 68.6% with BCI competition IV Datasets 2a, which is an encouraging result in motor imagery pattern recognition. And demonstrated that both-hands can be considered as one class in MI based BCI system.