Multi-Task Learning-Based Deep Neural Network for Steady-State Visual Evoked Potential-Based Brain-Computer...

Area of Science:

• Neuroscience
• Biomedical Engineering
• Artificial Intelligence

Background:

• Amyotrophic Lateral Sclerosis (ALS) severely impairs communication abilities.
• Brain-computer interfaces (BCIs) offer a potential communication pathway for individuals with severe motor impairments.
• Steady-state visual evoked potential (SSVEP)-based BCIs are a promising non-invasive technology.

Purpose of the Study:

• To develop a robust single-channel SSVEP-based BCI system for individuals with ALS.
• To enhance the practicality and accuracy of SSVEP-based BCIs through advanced machine learning techniques.
• To investigate the efficacy of multi-task learning for integrating denoising and classification in SSVEP-based BCIs.

Main Methods:

• Implementation of a single-channel SSVEP-based BCI system.
• Development of a neural network utilizing multi-task learning for simultaneous denoising and classification.
• Experimental validation of the proposed multi-task learning approach against other methods.

Main Results:

• The proposed multi-task learning approach effectively suppresses noise components in SSVEP signals.
• The system achieved high classification accuracy, demonstrating the integration of denoising and discriminative features.
• The multi-task learning strategy outperformed conventional approaches in SSVEP-based BCI performance.

Conclusions:

• Multi-task learning with denoising and classification is highly suitable for practical SSVEP-based BCI applications.
• This approach significantly improves the robustness and accuracy of BCIs for communication.
• Future work includes developing an augmentative and alternative communication interface for daily use by individuals with ALS.