Multiclass common spatial pattern with artifacts removal methodology for EEG signals

Common Spatial Pattern (SP) algorithm has been proved to be effective in Brain Computer Interface (BCI) for extracting features from Electroencephalogram (EEG) signals used in motor imagery tasks, but it is vulnerable to noise and the problem of over-fitting. Many algorithms have been devised to regularize CSP for two class problem, however they have not been effective when applied to multiclass CSP. The features extracted using the CSP are non-stationary in nature which increases the difficulty during classification. We propose a method to remove trials that are affected by noise before calculating the CSP. This helps in calculating eigenvectors which generates better CSP. To handle the non-stationarity in the EEG signal, Self-Regulated Interval Type-2 Neuro-Fuzzy Inference System (SRIT2NFIS) was proposed in the literature for two class EEG classification problem. This paper extends the SRIT2NFIS to Multiclass CSP using Joint Approximate Diagonalization (JAD). The results are presented on standard dataset.