Modeling a wireless network of sensors for Human Activity Recognition

Abstract

Much research has been on the study of human activity recognition. However, most of these studies use traditional methods focus on training and models optimization. Our search based on using data base adopted with radar sensors The radar data collected is first preprocessed and then converted into 2D images, providing information on frequency variation over time, also called micro-Doppler signature. These 2D images are then used to train deep learning algorithms to identify and classify different types of human activities. We chose Deep Learning algorithms for their ability to efficiently process complex data and for their flexibility in pattern recognition. In addition, we have implemented techniques to effectively manage data from multiple radars. changes have been made at database level as well as the new structure of models been established on the different databases established in order to target the processing the time of calculation and the precision thus their application in reality, we offer are adaptable and operational in real environments.