Gesture recognition on compact embedded devices is challenging due to limited processing power, memory, and storage. Most AI-based gesture detection systems rely on cloud processing or high-end hardware, making them unsuitable for real-time, low-power edge applications like wearables and smart devices.
This project demonstrates AI-powered gesture recognition running entirely on an edge device using the IndusBoard Coin. By leveraging Edge Impulse, a lightweight machine-learning model is trained to classify motion gestures using onboard sensor data and deployed directly on the microcontroller, enabling real-time gesture detection without cloud dependency.
Key Features
- On-device AI gesture recognition with no internet requirement
- Optimised ML model suitable for MCU-level constraints
- Uses built-in motion sensors for multi-axis gesture detection
- Real-time classification with serial output and action triggers
- Scalable to other sensor-based AI use cases
Applications
- Smartwatches and fitness trackers
- Touchless control for consumer electronics
- Industrial motion monitoring and anomaly detection
- Human–machine interaction systems
- Predictive maintenance and activity recognition
This project highlights how advanced AI capabilities can be successfully deployed on ultra-compact hardware like the IndusBoard Coin. By combining Edge Impulse with efficient sensor-driven models, it showcases a practical pathway for bringing intelligent gesture recognition to low-power, real-world edge devices.