Tracking and Identification of Targets via mmWave MIMO Radar

Sensing is essential to enable civil, industrial, and military applications that require situational awareness. Simultaneous tracking and identification of heterogeneous device-free targets (e.g., humans, robots, and vehicles) can provide information superiority for different types of operations and surveillance tasks. This paper presents a framework for tracking and identification of multiple device-free targets based on reflected radiofrequency signals. The proposed framework consists of (i) clutter mitigation and target detection relying on the estimated clutter intensity distribution in the environment; (ii) multitarget tracking relying on probabilistic data association; and (iii) neural network-based classification for target identification relying on time-domain representations of micro-Doppler signatures generated by target movements. We performed an experimentation, employing a frequency modulated continuous wave multiple-input-multiple-output radar at mmWaves, which validates the proposed framework. The experimental results, in terms of tracking and identification accuracies, show the benefits of using the proposed framework.