An Application of Artificial Intelligence to Adaptive Radar Detection Using Raw Data

In this paper, we address the detection of targets in clutter-dominated environments. Specifically, we devise and apply a new approach to solve the Interference Covariance Matrix (ICM) estimation problem based upon the neural networks. Assuming a specific structure for the ICM, we train a Neural Network (NN) to estimate the parameters that characterize the ICM in univocal way. Then, we use the results provided by the NN to build up an estimate of the entire ICM and plug it into the adaptive matched filter and the adaptive coherence estimator. The performance assessment is conducted by resorting to synthetic as well as real-recorded data and shows the effectiveness of the proposed approach also in comparison with conventional competitors relying on the sample estimates of the ICM parameters.