This is an AI-powered fall detection product developed as a final project for the Intelligent Interactive Products course, alongside an amazing team.
The product uses an IMU to track the real time position of an user and start a rescue protocol if a fall is detected. Said protocol is comprised of an alert to the emergency services, a visual countdown and a relaxing instructions regarding the actions that have to be taken by the fallen person.
The sensor device communicates via ESP Now to the base station, transmitting raw live accelerometer data to the base station, that is then transmitted to a computer running an AI model. Using a simple SVM, the fall is classified into multiple sub-types, each associated a corresponding feedback from the base device.
My task was the development of the embedded device, mainly the base station. I have worked on the code, electronics and designing and manufacturing the case.
The device uses an ESP32-C3 based board, an 8 LED RGB ring and a Bluetooth speaker that is connected to the main computer.
This acts as both a user interface, the LEDs showing the time that the emergency services would take to arrive, and the speaker offering clear and calming instructions for the person in need.
This project represented an exploratory opportunity of integrating machine learning algorithms in products, enabling new forms of intelligence and usefulness to emerge.
