Typical stochastic resonance models and their applications in steady-state visual evoked potential detection technology

The steady-state visual evoked potential (SSVEP) detection technology is more suitable for real-time brain -computer interface (BCI) systems due to its easy feature recognition, short response time, and high stability. Classic algorithms for SSVEP recognition such as the CCA and FBCCA methods require the construction of EEG templates and the spatial filtering of multi-electrode EEG data, which cannot obtain a satisfactory performance of SSVEP-EEG detection technology. To address this issue, we compared several stochastic resonance (SR) models in terms of potential functions and phase plane structures and utilized these models to enhance the amplitude and the energy of feature frequencies, thereby improving the recognition accuracy in SSVEP tasks. In our study, the SSVEP experiment recruited 34 subjects to evaluate and compare the validity of several SR models. Experimental results showed that SR-based methods can achieve higher detection accuracy and faster calculation speed compared with other traditional methods. Meanwhile, among these SR models, the underdamped second-order SR and FitzHugh-Nagumo SR have more outstanding characteristics for SSVEP-based BCIs.