Embedded Machine Learning System for Muscle Patterns Detection in a Patient with Shoulder Disarticulation

Area of Science:

• Biomedical Engineering
• Artificial Intelligence
• Rehabilitation Technology

Background:

• Advancements in artificial intelligence (AI) are crucial for developing sophisticated control systems for prosthetic devices.
• Electromyographic (EMG) signal classification is key to enabling intuitive and responsive prosthetic control.
• Existing systems often face limitations in portability and real-time processing capabilities.

Purpose of the Study:

• To develop and evaluate a portable AI-based system for real-time EMG signal classification.
• To enable actuator control for prosthetic devices using shoulder muscle activity.
• To enhance the autonomy and functionality of prosthetic limbs.

Main Methods:

• Utilized low-power microcontrollers for continuous EMG signal recording.
• Employed the Edge Impulse platform for AI model development and deployment on edge devices.
• Conducted a case study involving a patient with left shoulder disarticulation, collecting EMG data over two days for training and real-time testing.

Main Results:

• The AI system demonstrated accurate and swift real-time classification of EMG signals.
• The system successfully translated muscle electrical activity into control commands for prosthetic actuators.
• The portable design facilitated continuous signal recording and enhanced user mobility.

Conclusions:

• Portable AI-based systems show significant potential for advanced prosthetic control.
• Real-time EMG signal classification is feasible and can improve prosthetic user functionality.
• This technology offers a pathway to enhance the quality of life for individuals using prosthetic devices.