Multi-scale noise transfer and feature frequency detection in SSVEP based on FitzHugh-Nagumo neuron system

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Steady-state visual evoked potential (SSVEP) is a key brain-computer interface (BCI) control signal, valued for its robustness and minimal training needs.
• SSVEPs are often weak and contaminated by significant multi-scale noise, leading to poor signal-to-noise ratios and reduced performance with traditional linear methods.
• Existing template matching and spatial filtering algorithms struggle with the nonlinear and non-stationary characteristics of SSVEP under noise.

Purpose of the Study:

• To develop novel frameworks for extracting SSVEP features that overcome the limitations of linear methods in noisy environments.
• To improve the detection and classification accuracy of SSVEP signals for enhanced BCI performance.
• To investigate the efficacy of nonlinear dynamical systems, specifically the FitzHugh-Nagumo neuron model, for SSVEP signal processing.

Main Methods:

• Proposed two novel frameworks utilizing a two-dimensional nonlinear FitzHugh-Nagumo (FHN) neuron system for SSVEP feature extraction.
• Employed experimental validation with 22 human subjects to assess the proposed methods.
• Compared the performance of the FHN-based methods against traditional linear algorithms.

Main Results:

• The nonlinear FHN neuron model effectively transferred noise energy into the SSVEP signal, amplifying the target frequency amplitude.
• The proposed FHN and FHNCCA methods demonstrated higher classification accuracy compared to traditional approaches.
• Faster processing speeds were achieved with the FHN-based methods, leading to an improved information transmission rate for SSVEP-BCIs.

Conclusions:

• Nonlinear FHN neuron systems offer a promising approach for robust SSVEP extraction in noisy BCI applications.
• The developed FHN frameworks significantly enhance SSVEP detection performance, outperforming conventional linear methods.
• These advancements contribute to more effective and efficient SSVEP-based brain-computer interfaces.