Design & Implementation of a Real-time Image Processing Based Functional Prosthetic Arm Prototype

The increasing demand for affordable yet functionally advanced prosthetic limbs highlight limitations in traditional designs, such as restricted dexterity, unintuitive control, and high production costs. This paper presents a low-cost, vision-based prosthetic arm capable of real-time gesture recognition and multi-finger actuation. The system utilizes a USB webcam and Google MediaPipe for hand landmark detection, with gesture processing handled via Python. Encoded gesture data is transmitted to an Arduino Mega 2560, which controls five MG995R servo motors to actuate a tendon-driven mechanical hand, 3D-printed in ABS plastic. The arm replicates symbolic gestures including finger counting, the victory sign, and thumbs-up, while also supporting basic grasp types through independent finger control. The key novelty of this work lies in combining real-time gesture recognition with mechanically articulated actuation within a cost-effective and modular framework, making it suitable for assistive applications, academic research, and educational use. Simulation and hardware tests confirm accurate gesture-to-motion translation, low latency (~58 ms), and reliable grasping of common objects. Comparative benchmarking demonstrates superior modularity, responsiveness, and cost-efficiency over existing commercial solutions. The proposed system offers a viable foundation for assistive prosthetics, rehabilitation tools, and remote teleoperation in constrained environments.