Decoding Digital Visual Stimulation From Neural Manifold With Fuzzy Leaning on Cortical Oscillatory Dynamics

Area of Science:

• Neuroscience
• Computational Neuroscience
• Biomedical Engineering

Background:

• Decoding cognitive information from electroencephalogram (EEG) is vital for motion control and neural rehabilitation.
• High dimensionality and instability of EEG data hinder direct information extraction.
• Novel methods are needed to accurately interpret neural dynamics for practical applications.

Purpose of the Study:

• To propose a novel decoding method based on neural manifold of cortical activity for visual information.
• To identify critical visual information from EEG signals, particularly in the alpha frequency band.
• To enhance the accuracy of visual cognitive decoding for brain-computer interface (BCI) applications.

Main Methods:

• Designed visual experiments and analyzed EEG data across four major frequency bands.
• Employed manifold learning techniques, including t-distribution random adjacency embedded (t-SNE), to analyze alpha band activity.
• Developed a fuzzy system-based Takagi-Sugeno-Kang model using latent factors from t-SNE for EEG signal identification.

Main Results:

• EEG alpha band responses in frontal and occipital lobes were prominent for visual stimuli.
• t-SNE revealed cyclic neural manifold dynamics, with distinct loops for different tasks.
• The proposed method achieved 81.98% accuracy in visual cognitive decoding, improving to 83.05% with optimized features (frontal, parietal, occipital lobes).

Conclusions:

• Low-dimensional manifold dynamics offer a potential tool for decoding visual EEG signals.
• The combination of t-SNE and fuzzy learning significantly improves visual cognitive decoding accuracy.
• This approach has significant implications for BCI control, brain function research, and neural rehabilitation.