A Smartphone-Based sEMG Signal Analysis System for Human Action Recognition

Area of Science:

• Biomedical Engineering
• Rehabilitation Technology
• Human Action Recognition

Background:

• Surface electromyography (sEMG) provides objective data for lower-limb rehabilitation.
• Traditional sEMG analysis can be time-consuming and inconvenient.
• Need for accessible, real-time monitoring solutions in daily rehabilitation scenarios.

Purpose of the Study:

• To develop a portable sEMG acquisition device and mobile application for daily use.
• To enable real-time monitoring, analysis, and human action recognition (HAR) of lower-limb movements.
• To create a reliable system for clinical evaluation in lower-limb rehabilitation.

Main Methods:

• Development of a portable sEMG signal acquisition device and a companion mobile application.
• Collection of sEMG data for six distinct lower-limb rehabilitation exercises.
• Training a convolutional neural network (CNN) model using sEMG segments and action labels.

Main Results:

• The developed mobile application offers real-time sEMG data plotting, filtering, storage, and action recognition.
• The CNN model achieved high-precision human lower-limb action recognition, with a maximum accuracy of 97.96%.
• All tested lower-limb actions were recognized with over 97% accuracy.

Conclusions:

• The smartphone-based sEMG analysis system is effective for real-time monitoring and HAR in lower-limb rehabilitation.
• The system offers reliable data for objective clinical evaluation, improving rehabilitation efficiency.
• This technology facilitates convenient and accurate assessment of patient actions in daily settings.