STM32_NeuralNet_MovementDetection

Innovative project over neural networks and embedded systems. Check the report if you want to get extra informations and more precision about the results of this project.

Overview

This project aims at creating a Neural Network for STM32(L476RG) to do motion recognition with data from accelerometer. The neural net used is a multilayer perceptron and it has (currently) 3 layers.
The goals are :

• To use a neural network model (testing) on an stm32 to perform motion recognition
• To train the neural network on the stm32 itself
• To perform the testing algorithm on an FPGA to increase speed
• To compare performances between the stm32 alone and the stm32 with the fpga
• Check the report

Requirements

• STM32CubeMX : to edit project configuration (clock, gpios…)
• SystemWorkbench for STM32
• Accelerometer (ADXL345 used in this project)
• STM32 : STM32L476RG (Nucleo) used here

Credits

Four motivated Esisar students !

• Louka BARRIERE
• Gauthier CANET
• Romain LE DONGE
• Alexandre MORICEAU