A Novel Frequency Recognition Method for SSVEP Based on Multi-Stimulus Multi-Subset Strategy

Canonical correlation analysis (CCA) and its improved algorithms play an important role for SSVEP classification, mainly focusing on improving the signal-to-noise ratio by reducing the electroencephalogram (EEG) background noise. At present, the individual template strategy, the filter bank analysis strategy, the multi-stimulus analysis strategy and the extended analysis strategy are widely used in the CCA methods. This study proposes the multi-stimuli multi-subset extended canonical correlation analysis algorithm (msms-eCCA) for SSVEP recognition. The proposed method aims to exploit the positive effects of these strategies to improve the overall performance of the algorithm in terms of recognition accuracy and robustness. A comparative analysis is performed with four common state-of-the-art methods. Experimental results show that the proposed msms-eCCA method is superior to other algorithms in terms of recognition accuracy and information transfer rate. In particular, the msms-eCCA method shows superior performance at short test durations, achieving an accuracy of 86.46±2.91% and an information transfer rate of 168.34±8.02 bits/min at a test duration of 0.85 s. In addition, the msms-eCCA method shows improved resistance to interference.