TrueNorth-enabled real-time classification of EEG data for brain-computer interfacing.

Area of Science:

• Neuroscience
• Computer Science
• Biomedical Engineering

Background:

• Brain-computer interfaces (BCIs) offer potential solutions for individuals with locked-in syndrome.
• Existing BCI systems often require significant computational resources, limiting real-time applications on low-power platforms.

Purpose of the Study:

• To develop and evaluate a pipeline for real-time brain signal classification on a low-power neurosynaptic system.
• To assess the feasibility and accuracy of this pipeline for assisting individuals with communication and interaction.

Main Methods:

• Utilized electroencephalography (EEG) signals from a hand squeeze task.
• Implemented a convolutional neural network (CNN) with a time-preserving signal representation strategy.
• Deployed the pipeline on IBM's TrueNorth Neurosynaptic System for low-power, real-time processing.

Main Results:

• The proposed pipeline achieved a balance between high classification accuracy and real-time feasibility.
• Demonstrated successful real-time classification of brain signals on a low-power neuromorphic platform.

Conclusions:

• The developed BCI pipeline is effective for real-time brain signal classification on resource-constrained hardware.
• This approach shows promise for managing locked-in syndrome and can be adapted for conditions like spinal cord injury, epilepsy, and Parkinson's disease.