Spatio-Spectral Representation Learning for Electroencephalographic Gait-Pattern Classification.

Area of Science:

• Neuroscience
• Biomedical Engineering
• Machine Learning

Background:

• Human locomotion relies on complex brain coordination of muscles via the peripheral nervous system.
• Understanding gait mechanisms is crucial, but decoding cortical activity during varied walking conditions using EEG for gait classification remains underexplored.

Purpose of the Study:

• To develop and evaluate a novel deep neural network, spatio-spectral representation learning (SSRL), for classifying gait patterns from multi-channel EEG signals.
• To investigate the effectiveness of SSRL in decoding cortical processes across different walking conditions, including free and exoskeleton-assisted walking.

Main Methods:

• An EEG-based experiment was conducted involving four walking conditions: free walking and exoskeleton-assisted walking with varying assistive forces.
• A deep neural network topology, SSRL, with shared weights was proposed to learn spatial and spectral representations from EEG data.
• SSRL was compared against state-of-the-art methods for gait pattern classification.

Main Results:

• SSRL achieved a classification accuracy of 77.8%, surpassing existing methods in decoding gait patterns from EEG signals.
• Features extracted by SSRL were found to be more discriminative than hand-crafted features.
• Analysis of SSRL weights revealed spatial distributions consistent with known motor-activated cortical regions.

Conclusions:

• SSRL significantly advances the capability to decode human locomotion from EEG signals.
• The findings have potential implications for improving exoskeleton design, rehabilitation strategies, and clinical diagnostic tools for movement disorders.