A New Hybrid Method with Biomimetic Pattern Recognition and Sparse Representation for EEG Classification

This paper presents a novel classification framework combining Biomimetic Pattern Recognition (BPR) with Sparse Representation (SR) for Brain Computer Interface based on motor imagery. This framework can work well when encountering the overlap coverage problem of BPR by introducing the idea of SR. Using Common Spatial Pattern to extract the rhythm features of EEG data, we evaluate the performance of the proposed method in the datasets from previous BCI Competitions. By making comparison with those of LDA, SVM and original BPR, our proposed method shows the better classification accuracy.