A Low-rank Spatiotemporal based EEG Multi-Artifacts Cancellation Method for Enhanced ConvNet-DL's Motor Imagery
Summary
This study introduces a hybrid approach to remove artifacts from electroencephalograph (EEG) signals, significantly improving motor imagery (MI) decoding accuracy for brain-computer interfaces (BCI) in rehabilitation robots.
Area of Science:
- Neuroscience
- Biomedical Engineering
- Signal Processing
Background:
- Electroencephalograph (EEG) signals are crucial for decoding motor imagery (MI) for brain-computer interface (BCI) applications.
- Dynamic artifacts in EEG signals pose significant challenges for BCI system reliability in practical settings.
Purpose of the Study:
- To propose a hybrid approach for concurrent elimination of multiple EEG artifacts.
- To develop a deep learning model for enhanced MI task decoding from cleaned EEG signals.
Main Methods:
- A hybrid approach combining low-rank spatiotemporal filtering for artifact removal.
- Implementation of a convolutional neural network deep learning model (ConvNet-DL) for MI decoding.
- Comparative analysis with existing artifact removal techniques using EEG data from transhumeral amputees performing MI tasks.
Main Results:
- Significant improvements in MI task decoding accuracy (8.00–13.98%) and Mathew correlation coefficients (up to 14.38%) for the ConvNet-DL model.
- Achieved a signal-to-error ratio exceeding 11 dB.
- Demonstrated superior performance compared to existing artifact removal methods.
Conclusions:
- The proposed hybrid EEG artifact removal method combined with ConvNet-DL significantly enhances MI upper limb movement decoding accuracy.
- This approach offers potential for improved control input in BCI-based rehabilitation robotic systems.
