ResNet Applied for a Single-Snapshot DOA Estimation
Marcio L. Lima de Oliveira, Marco J.G. Bekooij · 2022 IEEE Radar Conference (RadarConf22) · 2022
In this paper, we discuss the usage of Residual Neural Networks (ResNets) for calculating the Direction Of Arrival (DOA) of MIMO radars and the estimation of the number of targets, using Minimum Redundancy Array (MRA). In addition, we are considering only the case of one snapshot at a time. This means that most techniques would deliver a poor estimation performance, whereas the Maximum Likelihood Estimation (MLE) algorithm delivers decent performance at a high computational cost. ResNet appears to be a viable alternative solution for this problem, being able to outperform MLE in some cases while having a less computational cost for scenarios with at least two different targets.