Group ensemble learning enhances the accuracy and convenience of SSVEP-based BCIs via exploiting inter-subject information

Over the past years, Steady-state visual evoked potential (SSVEP) is widely used in brain–computer interfaces (BCIs) due to its high efficiency, robustness and convenience. Although the ensemble task-related component analysis (eTRCA) is a state-of-the-art signal processing algorithm for SSVEP signals, the tedious training process restricts its extensive application. In this study, a novel hybrid framework is proposed to conquer this problem by reusing the validation data recorded previously. Additionally, a new group ensemble learning method is also proposed to combine a filter bank canonical correlation analysis (FBCCA) detector and several eTRCA detectors to achieve a high-performance SSVEP detector. The offline analysis, which used a four-class SSVEP dataset recorded from 9 subjects in four days, demonstrates that the proposed method outperforms the FBCCA method and almost achieves the performance of the eTRCA method. The proposed method provides a good trade-off between performance and training cost, which is more suitable for application in real life.