Eye-Gaze Controlled Wheelchair Based on Deep Learning.

Area of Science:

• Biomedical Engineering
• Artificial Intelligence
• Rehabilitation Technology

Background:

• Patients with Amyotrophic Lateral Sclerosis (ALS) often face progressive motor function loss, limiting their mobility and independence.
• Existing assistive technologies may not fully cater to the complex needs of ALS patients in natural environments.

Purpose of the Study:

• To design and develop an intelligent wheelchair system with eye-movement control for ALS patients.
• To enhance the natural environment navigation capabilities of assistive wheelchairs.
• To improve the smoothness and responsiveness of wheelchair motion.

Main Methods:

• Integration of an electric wheelchair, vision system, 2D robotic arm, and main control system.
• Utilizing a monocular camera to capture eye images for deep learning-based eye-movement direction recognition with an attention mechanism.
• Development of a motion acceleration model based on joystick trajectory and wheelchair speed to ensure smooth motion.
• Deployment of a lightweight eye-movement recognition model on an embedded AI controller.

Main Results:

• Achieved 98.49% accuracy in eye-movement direction recognition.
• Enabled wheelchair movement speeds of up to 1 m/s.
• Demonstrated smooth wheelchair movement trajectories without sudden accelerations.

Conclusions:

• The developed intelligent wheelchair system effectively utilizes eye-movement control for ALS patients.
• The system enhances mobility and independence in natural environments.
• The implemented motion acceleration model significantly improves the smoothness of wheelchair operation.