MACHINE LEARNING SYSTEM FOR PREDICTING ROBOTIC HAND FINGER MOVEMENTS USING A SMART GLOVE

Abstract

This research proposes a machine learning-based method to predict robotic hand finger movements using a Smart glove equipped with Light Dependent Resistor (LDR) sensors. Real-time finger flexion and extension data are captured via an ESP-WROOM-32 microcontroller interfaced through Arduino IDE and Jupyter. The multi-output regression problem of predicting simultaneous finger movements is addressed using a K-nearest neighbors regressor, optimized by Root Mean Square Prediction Error (RMSPE). Implemented in real-time, the system successfully synchronizes the robotic hand’s finger movements with the Smart glove inputs, enhancing control precision, reducing latency, and improving user experience. This approach holds promise for advancing artificial limb control and remote robotic surgery applications.