Synapse : The NeuroTech Challenge

Surface Electromyography (sEMG) signals carry information about muscle activation patterns that precede limb motion. By decoding these biosignals, one can infer and control prosthetic limb movements using machine-learning models. Participants will use a multichannel EMG signals recorded during different arm or finger gestures to train a machine learning model that maps muscle electrical activity to prosthetic control outputs such as: Open Hand, Closed Hand, Lateral Pinch, Signalling Sign, Rock Sign. The machine learning–based control model must interpret multichannel EMG signals from forearm muscles to predict discrete gesture actions. The Problem Statement: Design, train, and evaluate a supervised classification model that predicts discrete gestures. The final system should demonstrate real-time prosthetic control simulation and map gestures to a simple gaming interface. Dataset Description: The dataset contains surface electromyography (sEMG) data of 5 different hand gestures performed by six subjects. The dataset will be released on the portal on 10th January, 2026. Phases: Model Development and Submission (Online): Participants submit their trained model predictions on a test dataset whose ground truth is withheld. Live Demonstration - Game Interface (top 5 teams): Selected teams attach 4 sEMG channels on their forearm. Real-time EMG is streamed to their trained model. A general game interface will be provided for any participant or visitor to connect sensors and map gestures to actions like fire/reload.