Assisting Upper Limb Prosthesis with a Computer Vision System for Material Detection.

Area of Science:

• Biomedical Engineering
• Computer Vision
• Robotics

Background:

• Computer vision systems struggle with noisy backgrounds in real-world prosthetic control.
• Myoelectric prostheses require advanced control strategies for improved functionality.

Purpose of the Study:

• To assess the feasibility of using YOLOv8 for object identification in myoelectric prosthetic hand grasping tasks.
• To integrate a custom object detection model with an EMG-based prosthesis controller.

Main Methods:

• A state-of-the-art object detection model (YOLOv8) was trained to recognize various types of cups.
• The custom YOLOv8 model was integrated with an EMG-based prosthetic hand controller.
• A human subject performed object grasping tasks to evaluate model performance.

Main Results:

• The custom YOLOv8 model achieved approximately 93% classification accuracy in identifying grasped objects.
• The model demonstrated robust performance during object grasping tasks.

Conclusions:

• Object detection using YOLOv8 is feasible for myoelectric prosthetic hand control.
• Future work will utilize object recognition to dynamically adjust grip force parameters for enhanced prosthetic function.