Electromyographic Signal Analysis for Monitoring and Classifying Hand Muscle Fitness

---

Abstract

In fields like sports science, rehabilitation, and ergonomics, hand muscle fitness is a crucial measure of general physical health. One popular non-invasive technique for capturing muscle activity during contraction is electromyography, or EMG. But it is hard to look at raw EMG signals because they have noise and their signals change. To solve this problem, advanced signal processing and machine learning methods are used to find features and group muscle strength. This paper presents a method for classifying hand muscle strength (Low, Medium, High) from Electromyography (EMG) signals using Convolutional Neural Networks (CNNs) and Multi- Layer Perceptron Model (MLP). EMG signals acquired using the Muscle BioAmp Patchy sensor, were transformed into frequency domain using Fast Fourier Transform (FFT). Four key frequency-domain features Mean Frequency, Peak Frequency, Spectral Frequency and Spectral Entropy were extracted. K- means clustering was initially applied to these features to identify potential groupings. Two supervised learning models were trained using the clustered data: a Convolutional Neural Network (CNN) and a Multi-Layer Perceptron (MLP). The clustered data was then used as input to a CNN model. The CNN effectively learned to classify muscle strength, achieving an overall accuracy of 100%. This approach demonstrates the potential of combining FFT-based feature extraction, K-means clustering and CNNs for accurate and automated muscle fitness assessment. The MLP model, while simpler in architecture, also performed well with an accuracy of 88.46%, making it suitable for applications with limited computational resources. This dual-model approach demonstrates the potential of combining FFT-based feature extraction, unsupervised clustering, and deep learning for accurate, efficient, and automated assessment of muscle strength.

---

Keywords

Electromyography, Fast Fourier Transform (FFT), Feature Extraction, K-means Clustering, Convolutional Neural Networks (CNNs), Multi Laper Perceptron Model (MLP), Machine Learning, Muscle Strength Classification