Improvement of Classification Accuracy in a Phase-Tagged Steady-State Visual Evoked Potential-Based Brain-Computer Interface Using Adaptive Neuron-Fuzzy Classifier

Steady-state visual evoked potential (SSVEP) has been used to design brain-computer interface (BCI) for a variety of applications, due to its advantages of high accuracy, fewer electrodes, and high information transfer rate. In recent years, researchers developed phase-tagged SSVEP-based BCI to overcome the problem of amplitude-frequency preference in traditional frequency-coded SSVEPs. However, the phase of SSVEP could be affected by subject's attention and emotion, which sometimes causes ambiguity in discerning gazed targets when fixed phase margins were used for class classification. In this study, we adopted adaptive neuron-fuzzy classifier (ANFC) to improve the gaze-target detections. The SSVEP features in polar coordinates were first transformed into Cartesian coordinates, and then ANFC was utilized to improve the accuracy of gazed-target detections. The proposed ANFC-based approach has achieved 63.07 +/- 8.13 bits/min.