Scholarlink

This study presents a machine learning approach to classify electromyography (EMG) signals recorded from lower limb muscles. It focuses on using these signals to distinguish between different physical activities and physiological conditions. The main goal is to differentiate between normal and abnormal muscle conditions and to classify activities such as sitting, standing, and walking based on EMG signal features. This research uses both Random Forest (RF) and Support Vector Machine (SVM) models to examine and compare how accurately these algorithms classify muscle activity under various conditions. In the two classifications, we find that Random Forest classification achieves an accuracy of 73%, while Support Vector Machine reaches 58%. This shows that RF is much more effective for this type of nonlinear data. Our findings indicate that RF handles nonlinear correlations in EMG signals better than SVM. The study also addresses issues like subject variability, noise interference, and real-time applicability. Future research will focus on increasing dataset diversity and exploring real-time classification methods.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

EMG Signal Classification in Lower Limbs Using Machine Learning Techniques

  • 章节

  • 发表日期：

    2026年

查看全文
  • Vivek Saha,
  • Sivakumar Rajagopal

摘要

This study presents a machine learning approach to classify electromyography (EMG) signals recorded from lower limb muscles. It focuses on using these signals to distinguish between different physical activities and physiological conditions. The main goal is to differentiate between normal and abnormal muscle conditions and to classify activities such as sitting, standing, and walking based on EMG signal features. This research uses both Random Forest (RF) and Support Vector Machine (SVM) models to examine and compare how accurately these algorithms classify muscle activity under various conditions. In the two classifications, we find that Random Forest classification achieves an accuracy of 73%, while Support Vector Machine reaches 58%. This shows that RF is much more effective for this type of nonlinear data. Our findings indicate that RF handles nonlinear correlations in EMG signals better than SVM. The study also addresses issues like subject variability, noise interference, and real-time applicability. Future research will focus on increasing dataset diversity and exploring real-time classification methods.