Machine Learning Models for Indoor Person Monitoring using Radar Signals (PhD)

This thesis aims to exploit a commercial frequency modulated millimetre wave (FMCW) radar and machine learning algorithms to recognise human indoor activity which is a principal feature in many context like patient monitoring and smart homes. This radar operates in the frequency range of 60 Ghz and 64 Ghz and can get the parameters range, velocity and position of each object in its field of view. By using a set of signal processing algorithms related to radar signals, the result is a set of features fed to a machine learning learning algorithm for human activity recognition. This algorithm could be a recurrent convolutional neural network (R-CNN) or a CNN or a simple classifier, the choice would be based on the extracted features.