A Novel Multiple Motor Imagery Experimental Paradigm Design and Neural Decoding

Motor imagery (MI) is a typical active brain-computer interface (BCI), which can be decoded by the pattern of EEG when the subject conduct motor imagery (e.g. left hand, right hand, feet and tongue.). In this paper, we proposed a novel multiple MI experimental paradigm which includes left hand, right hand, both hands, and feet. Based on the proposed paradigm, ten subjects participated in the data collection experiment at Jiangmen Brain-like Computation and Hybrid Intelligence R&D Center. Furthermore, we proposed an algorithm for MI neural decoding. Firstly, combining the one-vs-rest (OVR) strategy with the filter bank common spatial pattern (FB-CSP), we extract the features of multiple-class MI. Secondly, F-score optimization is used for features selection and then features are feed to a spiking neural network (SNN) for classification. Evaluation experimental results show that the averaged accuracy for four MI classification was up to 61.30% (average kappa: 0.482). The feasibility and effectiveness of the experimental paradigm prove that it can be used as an alternative way and has practical significance.