Adaptive Detection of Frequency-Hopping Spread Spectrum Signals Based on Compressed Measurements and Artificial Neural Networks

In modern secure communications, the frequency-hopping spread spectrum (FHSS) technique has been widely implemented. In this technique, the signal carrier frequency hops in a pseudo-random manner, and the conventional non-cooperative receivers without the knowledge of the hopping sequence have to sample the signals with relatively expensive hardware to catch the entire spectrum. Thus, the FHSS signals usually get good performance against interception. In this paper, the authors propose a non-cooperative method to detect the FHSS signals from compressive measurements. In contrast to most of the existing literature on compressive sampling, which proposed to use random or recursively optimized measurement kernels, the proposed method enable efficient adaptive measurement kernel design based on task-specific information optimization and artificial neural networks. Simulation results demonstrate that the proposed method provides significantly improved detection accuracy over the compressive detection method with random kernels, as well as significantly improved measurement kernel designing efficiency over the recursive optimization method in the existing literature.