Overview
The F.A.S.T. wearable is designed to provide boxers and martial artists with real-time feedback on their punching technique. By integrating multiple sensors and advanced data processing, it captures force, speed, and trajectory data to analyze punch effectiveness and form.
This project demonstrates end-to-end hardware development, from mechanical design through PCB layout, firmware development, and machine learning classification.
Technical Implementation
- Mechanical Design: Custom housing and wrist-mounted enclosure designed in SolidWorks, optimized for comfort and sensor placement
- PCB Design: Two-sided custom PCB in EasyEDA, optimized for size and commercial production readiness
- Hardware: ATmega328 MCU, MPU-6500 IMU for motion tracking, capacitance-to-digital load cell for force measurement, USB-C interface via FT231XQ
- Assembly: Hand-assembled SMT board using stencil and solder paste; custom 3D-printed jig for second-side reflow
- Firmware: C/C++ code integrating IMU data using Euler's method for position derivation from acceleration
- Data Processing: Python-based random forest classifier trained on IMU signals to identify punch types (jab, cross, hook, uppercut)
Key Challenges & Solutions
- Miniaturization: Balancing sensor accuracy with wearable form factor required careful PCB layout and component selection
- Data Integration: Combining accelerometer, gyroscope, and force sensor data into meaningful punch classification
- Real-time Processing: Implementing efficient algorithms on resource-constrained microcontroller
- Calibration: Developing calibration procedures for consistent force and motion measurements
Results & Impact
The wearable successfully captures and classifies different punch types with over 90% accuracy. The compact design allows for practical use during training sessions, providing immediate feedback to athletes.
This project showcases the integration of mechanical design, electronics, embedded programming, and machine learning in a consumer-ready product concept.