Master thesis : Embedded Neural Network for Movement classification and Fall Detection

Abstract

With the the evolution of ARM-based SoC and the development of dedicated libraries for intelligent system on microcontrollers, more than ever we have the possibility of creating powerful, power efficient and cheap connected devices with the capability of solving a wide range of issues. The goal of this work is to explore the possibility of running a neural network model on a connected bracelet with the higher intent of being able to identify and classify falls. This thesis subject was proposed and supported by IoT-D, a young company focused on innovating in the field of connected devices. It centers around the Iomoov, a connected bracelet of their design. We conclude this thesis by suggesting multiples clues for improvement as well as explore the limitations faced during the work and how to address them.