ReachAI: Machine Learning-Based Prosthetic Hand for Adaptive Rock Climbing

Overview

Rock climbing is inaccessible to individuals with hand amputations because existing myoelectric prosthetics cannot perform sport-specific grips like crimps, pinches, and slopers. This project develops an electromyography (EMG) controlled prosthetic hand that recognizes and executes seven climbing grips from forearm muscle signals, demonstrating the feasibility of adaptive climbing technology for limb-different athletes. Using the publicly available NinaPro Database 2 (39 subjects, 12-channel forearm EMG at 2000 Hz), I extracted 6 physiological measurements (Mean Absolute Value, Root Mean Square, Waveform Length, Zero Crossings, Mean and Median Frequency) from 12 different EMG channels, totaling 72 data points, from 100ms sliding windows with 75% overlap, enabling classifications every 25ms. A Random Forest classifier with 200 decision trees was trained on 38 subjects using subject-independent methodology and tested on one separate subject to evaluate generalization to new users, a critical requirement for prosthetic deployment. The classifier identifies seven grips: open hand, jug, crimp, half-crimp, pinch, sloper, and pocket. The system runs on a MacBook Air and sends commands via USB serial at 115200 baud to an Arduino UNO Q, which controls five servos in a Lewan Soul robotic hand using pre-calibrated positions, achieving 150ms total system latency suitable for real-time prosthetic control. The system achieved 83% subject-independent classification accuracy with 25ms processing time. Biomechanically distinct grips showed excellent results (OPEN: 100% recall, PINCH: 79%, JUG: 76%), while similar grips like CRIMP (12%) and HALFCRIMP (0.5%) were confused due to overlapping muscle activation patterns, reflecting known limitations in surface EMG pattern recognition. The 83% accuracy compares favorably to published myoelectric prosthetic studies (Atzori et al.: 60-70%), demonstrating potential for sport-specific prosthetics and future applications in adaptive sports rehabilitation and assistive technology.

Competition history

• CSEF 2026 Medicine & Physiology (Senior Division) · Entry S-15-18

Related projects

Closest projects by meaning, across every fair and year in the corpus.

Browse more like this

• All Biomedical and Health Sciences projects
• CSEF Biomedical and Health Sciences projects
• Every CSEF 2026 project