Classifying Hand Gestures through EMG Data with Machine Learning

This study explores the application of machine learning techniques in the classification of hand gestures for controlling upper limb neuroprosthetics. The non-stationary nature of sEMG in practical applications presents a challenge in maintaining consistent performance. To address this issue, we applied five distinct machine learning methods to classify hand gesture data collected from 8 sensors through a Myo wristband. The Kaggle dataset provided over 2900 samples from 4 classes, resulting in over 2900 samples per group. We evaluated the performance of the classification methods based on various metrics such as precision, recall, F1 score, support, and accuracy. Our findings show that QDA performed the best due to its high accuracy. The study identified and classified four hand gestures based on their performance metrics. Our research provides a simple approach to controlling smart equipment like prosthetic limbs, wheelchairs, entertainment software, and human-machine interaction. The use of machine learning in the classification of hand gestures for neuroprosthetics has the potential to significantly enhance the functionality and ease of use of these devices.