CCA Model with Training Approach to Improve Recognition Rate of SSVEP in Real Time

Brain Computer Interfaces (BCIs) are often used to control external devices using electroencephalogram (EEG) signals. In Steady-State Visually Evoked Potentials (SSVEP) based BCIs, suboptimal Information Transfer Rate (ITR) is achieved due to false detection of SSVEP as one of the target class while the subject is not focusing on any target. To alleviate this issue, we propose a class labelling method where a classifier is trained against the non-target class. In the experiment, features are extracted using Canonical Correlation Analysis (CCA) and class labelling is performed using the proposed method. Afterwards, Linear Discriminant Analysis (LDA) has been employed for classification task. The results were compared with standard methods such as CCA and Fast Fourier Transform (FFT), implemented for the same experimental setup. The proposed method was found to be highly accurate and it successfully overcame the issues found in previous methods.