Project Demonstration
Advanced EMG-Controlled Prosthetic Hand | Harvard Engineering Sciences 50
Revolutionary low-cost, real-time trainable bionic hand using EMG signals to control a 3D-printed prosthetic. Demonstrates circuit-based signal preprocessing for enhanced myoelectric prosthetic accuracy.
The circuit design centers on advanced EMG signal acquisition and processing, utilizing specialized amplification circuits to capture muscle electrical activity with high fidelity. The system employs custom circuit preprocessing to filter and amplify EMG signals before digital conversion, ensuring clean signal input for our machine learning algorithms. The complete hardware implementation costs $121.26, demonstrating that sophisticated myoelectric control can be achieved through cost-effective component selection and efficient circuit design.
The prosthetic hand design incorporates servo-controlled finger mechanisms capable of binary classification for open and close gestures. Our kNN algorithm implementation in C++ processes EMG signals in real-time, enabling sub-minute training periods that allow users to quickly calibrate the system to their specific muscle patterns. The algorithm analyzes muscle contraction patterns and translates them into precise hand movements, creating an intuitive control interface that responds immediately to user intent.
Leading the research team of Finn Seyffer, Xavier Ayala-Vermont, and Anthony Bynum, I was responsible for conceptual planning, circuit assembly, EMG amplification design, and the binary classification implementation. The system achieves reliable real-time EMG processing through optimized signal processing pipelines and efficient algorithm execution. This research demonstrates how advanced biomedical engineering principles can be applied to create functional myoelectric prosthetics that bridge the gap between biological signals and mechanical control systems.