Learning Scanning Regime For Electronic Support Receivers by Nonnegative Matrix Factorization

Abstract

Narrow-band receivers used in electronic support systems should operate with a frequency scanning strategy in order to detect radar signals in different frequency ranges of the electromagnetic spectrum. This scanning strategy can be determined with learning-based models in an environment where the parameters of the radars are unrecognized. In previous studies, the problem is modeled as a dynamic system with Predictive State Representations and the resulting optimization problem is solved via Singular Value Thresholding (SVT) algorithm. We propose a scanning regime learning method based on Nonnegative Matrix Factorization (NMF) algorithm. The proposed method requires less computation time for subspace identification in each iteration. According to the simulation results, the average calculation time is reduced around 40% by using NMF without any loss of detection performance.