MuscleNET: Smart Predictive Analysis for Muscular Activity Using Wearable Sensors

Abstract

Doing weightlifting training at home has become more popular during the pandemic. Unfortunately, exercising without professional help can lead to dangerous injuries such as muscle tearing. It is possible to create a smart system with machine learning to overcome muscle injuries and suggest an appropriate training program. The use of suitable algorithms enables us to develop programs that can perform predictions based on sEMG (Surface Electromyography) signals. In this study, sEMG signals are collected from the skin surface and features are extracted to be used in deep learning networks. A wearable hardware collects sEMG signals and transfers them to our mobile application via Bluetooth. The mobile application transfers data to the cloud to make predictions based on sEMG signals. We developed MuscleNET for training monitoring, injury prediction/detection, and training quality prediction. Initial measurements indicate that MuscleNET can be used effectively for training quality prediction and real time training monitoring.