A 3D Printed, Bionic Hand Powered by EMG Signals and Controlled by an Online Neural Network.

Area of Science:

• Biomedical Engineering
• Artificial Intelligence
• Rehabilitation Technology

Background:

• Approximately 8% of Ecuador's population experiences limb amputation, facing significant employment barriers due to high prosthesis costs and low average wages.
• Limited access to affordable and functional prosthetic devices exacerbates the labor disadvantage for amputees, with only 17% employed.

Purpose of the Study:

• To design and develop an economically accessible 3D-printed hand prosthesis.
• To enable real-time control of the prosthesis using electromyography (EMG) signals and artificial intelligence (AI).

Main Methods:

• Developed an experimental methodology to record upper extremity muscle activity using three surface EMG sensors for specific tasks.
• Trained a five-layer neural network using recorded EMG data and compressed the model using TensorflowLite.
• Designed a 3D-printed hand prosthesis with a gripper and pivot base using Fusion 360, integrating an ESP32 development board for real-time EMG signal processing and actuation.

Main Results:

• Created a database of 60 electromyographic activity records from three distinct muscle tasks.
• Achieved 78.67% accuracy in classifying muscle tasks with an 80 ms response time.
• The 3D-printed prosthesis demonstrated the capacity to support 500 g with a safety factor of 15.

Conclusions:

• The developed system offers a viable, low-cost solution for prosthetic hand control.
• Integration of EMG signals and AI in 3D-printed prosthetics can significantly improve functionality and accessibility for amputees.
• This technology has the potential to reduce the labor disadvantage faced by individuals with limb amputations.