Deep Learning-Based Implementation of Telerobotic Control with Gesture Recognition from EMG Signals

This research explores the field of telerobotics, focusing on the deep learning based remote control of a robot using Electromyography (EMG) and an Inertial Measurement Unit (IMU). The aim is to enable control of a robot arm by capturing muscle responses via electrical activity using the Myo armband. The captured signals representing hand gestures are classified using a four layer wide fully connected neural network, allowing for seamless translation of gestures into robot commands. Training the model with online and self-collected datasets demonstrates the effectiveness of the proposed method. The model has an accuracy of 82.92% on previously unseen data. Successful implementation of telerobotics relies on the ability of the Myo armband to capture EMG signals and IMU data, facilitating control over the end effector of the robot and movement through gesture recognition and orientation conversion.