Towards Intuitive Human-Robot Interaction: A Machine Learning Approach to Gesture Recognition

This work presents a human gesture recognition system using surface electromyography signals (sEMG) in order to improve human-robot interaction (HRI). A cobot implementation via simulation in RoboDK is also proposed. A structured experimental protocol is proposed that optimizes the acquisition of sEMG and EEG signals and minimizes noise to obtain high-fidelity data before preprocessing. Three hand and arm gestures were defined to establish interaction with the robot, such as Ask, Take, and Neutral. Three machine learning models were trained and tested with raw sEMG data and preprocessed sEMG data, surpassing 96% recognition accuracy using the Bagged Trees model with preprocessed sEMG data. This approach offers promising applications in assistive robotics, rehabilitation, and occupational ergonomics. In addition, the study lays the foundation for future developments that integrate EEG signals, with the aim of improving the accuracy of the system and allowing recognition of movement intentions.