Deep Learning-Based Radar Processing: Simultaneous Target Classification, Activity Detection, and Range Estimation with Micro-Doppler Radar

This research investigates the application of deep learning techniques for executing multiple tasks within a single neural network, specifically focusing on the simultaneous classification of radar targets, detection of target activities, and determination of target range. The classification process utilizes a micro-Doppler radar operating in the millimeter-wave regime, which provides complete clutter immunity and the ability to distinguish even the most subtle target movements. By integrating millimeter-wave radar with a neural network, a sophisticated system is developed that can differentiate between humans and animals, identify the presence of a weapon carried by an individual, and ensure immunity from deception attempts, such as a person crawling and pretending to be a walking animal. In addition, since the radar used in this study does not inherently provide range detection capabilities, the neural network is responsible for extracting range information from the data. The objective of this study was to gain a better understanding of the neural network's decision-making process and the mechanisms by which it is able to perform multiple tasks at once by conducting an explainable AI analysis. This analysis offers valuable insights into the inner workings of the neural network across various tasks, highlighting the potential for further development and optimization of this deep learning-based radar processing approach.