Electromyography-Based Hand Pose Estimation U sing Machine Learning

This paper presents an approach to hand pose estimation using Electromyography (EMG) signals with machine learning techniques. Long Short-Term Memory (LSTM) was shown to be the most accurate neural network architecture for identifying hand gestures from EMG signals. The EMG data were gathered from 4 classes of 15 participants, each of whom made a gesture five times for every class. T he model's output indicates an accuracy of roughly 93 %, which makes it feasible for hand posture estimation. Such technology plays a pivotal role in various applications, including human-computer interaction, re-habilitation, prosthetics, virtual reality, and gaming. Leveraging EM G signals offers a direct and intuitive means of inferring hand gestures and movements and better protects the user's identity than other motion capture techniques such as video images. This research has the potential to contribute to the growing field of human-machine interaction and assistive technology by offering a reliable and efficient means of estimating hand poses.