Embedded Machine Learning Using a Multi-Thread Algorithm on a Raspberry Pi Platform to Improve Prosthetic Hand...

Area of Science:

• Biomedical Engineering
• Machine Learning
• Embedded Systems

Background:

• Developing advanced prosthetic hands requires high accuracy and real-time operation.
• Existing systems often face challenges with computational efficiency and responsiveness.

Purpose of the Study:

• To develop and evaluate a real-time embedded system for prosthetic hand control.
• To implement time-domain feature extraction and machine learning on a System on Chip (SoC) using a multi-thread algorithm.

Main Methods:

• Collected EMG signals from ten healthy volunteers using dry electrodes on wrist muscles.
• Applied four time-domain features for EMG signal extraction.
• Evaluated k-nearest neighbor (k-NN), Naive Bayes (NB), decision tree (DT), and support vector machine (SVM) algorithms.
• Implemented data acquisition, feature extraction, machine learning, and motor control using a multi-thread algorithm on a Raspberry Pi SoC.

Main Results:

• The combination of Mean Absolute Value (MAV) feature and decision tree (DT) machine learning achieved the highest accuracy (98.41%).
• The system demonstrated a computation time of approximately 1 ms.
• The multi-thread algorithm implementation significantly reduced processing time.

Conclusions:

• Multi-thread algorithm implementation in pattern recognition enhances accuracy and decreases computation time for prosthetic hand control.
• The developed system offers a promising real-time solution for prosthetic hand operation.