A lightweight spiking neural network for EEG-based motor imagery classification.

Area of Science:

• Neuroscience
• Artificial Intelligence
• Signal Processing

Background:

• Spiking neural networks (SNNs) mimic the human brain's neural processing using sparse spike events.
• SNNs offer potential for effective and energy-efficient spatio-temporal signal processing.
• Motor imagery (MI) classification using electroencephalogram (EEG) is a key brain-computer interface (BCI) application.

Purpose of the Study:

• To propose a lightweight SNN model for EEG-based MI classification.
• To develop a model with brain-inspired architecture, energy efficiency, and dataset agnosticism.
• To evaluate the SNN model's performance against established deep learning methods.

Main Methods:

• A novel, lightweight SNN model was designed.
• The model incorporates a brain-inspired architecture.
• Experiments were conducted on three public EEG datasets for within-subject and cross-subject MI classification.

Main Results:

• The proposed SNN model demonstrated superior performance in EEG-based MI classification.
• The SNN model outperformed four classical convolutional neural network (CNN) based models.
• The model's effectiveness was validated across different experimental conditions and datasets.

Conclusions:

• The developed lightweight SNN model is highly effective for EEG-based MI classification.
• This SNN approach offers an energy-efficient and adaptable solution for BCIs.
• The findings suggest SNNs are a promising alternative to CNNs for BCI applications.