Weak Feature Extraction and Strong Noise Suppression for SSVEP-EEG Based on Chaotic Detection Technology

Area of Science:

• Neuroscience and Biomedical Engineering
• Signal Processing and Machine Learning

Background:

• Brain-computer interfaces (BCI) offer alternative communication by translating brain activity into commands.
• Steady-state visual evoked potential (SSVEP) is a common BCI paradigm, but its weak EEG signals are susceptible to noise.
• Traditional SSVEP classification methods struggle with feature extraction, particularly for BCI-Illiterate users.

Purpose of the Study:

• To introduce a novel Chaos theory-based method for SSVEP feature extraction.
• To improve BCI performance, especially for individuals with weak EEG responses (BCI-Illiteracy).
• To evaluate the effectiveness of the proposed Chaos theory method against traditional algorithms.

Main Methods:

• A new method utilizing nonlinear dynamics and Chaos theory to detect SSVEP features by analyzing chaotic system state changes.
• Recruitment of 32 subjects, divided into normal (Group A) and BCI-Illiterate (Group B) based on preliminary classification accuracy.
• Comparative analysis of classification accuracy and information transmission rates using Chaos theory versus traditional methods.

Main Results:

• All tested classification methods performed well for normal subjects.
• The Chaos theory-based method demonstrated excellent performance and significant improvements for BCI-Illiterate subjects.
• This approach effectively addresses the challenge of weak SSVEP signals and multi-scale noise in BCI applications.

Conclusions:

• Chaos theory offers a powerful new approach for SSVEP feature extraction in BCI.
• The proposed method significantly enhances BCI performance for BCI-Illiterate individuals.
• This research paves the way for more robust and accessible brain-computer interfaces.