Frequency Domain Approach in CSP based Feature Extraction for EEG Signal Classification

EEG signal in the time domain with high sampled rate faces difficulties for their noise sensitive properties that lead to erroneous feature extraction. Since the feature extraction is one of the most significant steps in EEG signal classification, common spatial pattern (CSP) is a widely used approach for feature extraction. Conventional CSP in the time domain may often fail to maintain the discriminative features between the classes. Therefore, a frequency domain CSP (FCSP) is proposed by the work to overcome the limitations of the conventional CSP. We have applied the conventional and FCSP method on the motor imagery data for feature extraction. The average classification accuracies of the conventional and FCSP method were found 74% and 84%, respectively. Eventually, the proposed scheme outperforms the conventional method by increasing the classification accuracy up to 10%.