Evaluating and Designing Machine Learning Models to Classify Gestures from Electromyography (EMG) data
Session Number
3
Advisor(s)
Dr. Ashwin Mohan, IMSA
Location
A123
Discipline
Computer Science
Start Date
15-4-2026 2:15 PM
End Date
15-4-2026 3:00 PM
Abstract
The aim of this study was to evaluate and design machine learning models to perform gesture classification on Electromyography(EMG) signals. EMG signals are the electrical activity produced by neurons to stimulate and manipulate muscles. These signals have become a central modality for decoding human motor intent in applications such as prosthetic control and human-machine interaction. However, these signals are inherently noisy, non-stationary, and subject-dependent, which makes effective signal processing and feature extraction a challenge. Machine learning models have been employed as a powerful method to process these signals, as models excel at pattern recognition. In this research, we extracted Time-Domain and Frequency-Domain features and used them to train various machine learning models, such as: k-Nearest Neighbors (KNN), Support Vector Machine (SVM), Random Forest (RF), Logistic Regression (LR), Naïve Bayes (NB), Multilayer Perceptron (MLP). We used the NinaProDB2 dataset where subjects performed various gestures and instructed the ML models to classify EMG signals into respective gestures.
Evaluating and Designing Machine Learning Models to Classify Gestures from Electromyography (EMG) data
A123
The aim of this study was to evaluate and design machine learning models to perform gesture classification on Electromyography(EMG) signals. EMG signals are the electrical activity produced by neurons to stimulate and manipulate muscles. These signals have become a central modality for decoding human motor intent in applications such as prosthetic control and human-machine interaction. However, these signals are inherently noisy, non-stationary, and subject-dependent, which makes effective signal processing and feature extraction a challenge. Machine learning models have been employed as a powerful method to process these signals, as models excel at pattern recognition. In this research, we extracted Time-Domain and Frequency-Domain features and used them to train various machine learning models, such as: k-Nearest Neighbors (KNN), Support Vector Machine (SVM), Random Forest (RF), Logistic Regression (LR), Naïve Bayes (NB), Multilayer Perceptron (MLP). We used the NinaProDB2 dataset where subjects performed various gestures and instructed the ML models to classify EMG signals into respective gestures.