Five-Class SSVEP Response Detection using Common-Spatial Pattern (CSP)-SVM Approach

Abstract

Brain-computer interface (BCI) technologies significantly facilitate the interaction between physically impaired people and their surroundings. In electroencephalography (EEG) based BCIs, a variety of physiological responses including P300, motor imagery, movement-related potential, steady-state visual evoked potential (SSVEP) and slow cortical potential have been utilized. Because of the superior signal-to-noise ratio (SNR) together with quicker information transfer rate (ITR), the intentness of SSVEP-based BCIs is progressing significantly. This paper represents the feature extraction and classification frameworks to detect five classes EEG-SSVEP responses. The common-spatial pattern (CSP) has been employed to extract the features from SSVEP responses and these features have been classified through the support vector machine (SVM). The proposed architecture has achieved the highest classification accuracy of 88.3%. The experimental result proves that the proposed architecture could be utilized for the detection of SSVEP responses to develop any BCI applications. Keywords: EEG, BCI, SSVEP, CSP, SVM, Machine Learning