Final Year Project

Project Title

FPGA-Accelerated Transformer for sEMG-Based Gesture Prediction

Description

This project addresses the accessibility gap for advanced prosthetics in Pakistan by developing an affordable, EMG-based control system. The aim is to bridge the divide between rudimentary mechanical devices and high-cost imported myoelectric hands. By leveraging a lightweight, quantized Transformer model, the system tracks continuous hand motion from non-invasive surface EMG signals. To achieve real-time performance, the model is implemented on a low-cost FPGA, utilizing its inherent parallelism for high-speed, power-efficient processing. The purpose is to provide a "brain" for prosthetics that translates muscle activity into accurate joint angles for virtual or physical hands with low latency. This hardware-software co-design ensures high-accuracy gesture recognition on a portable edge device.