Weak Signal Detection Based on Beta Divergence

Among different electromagnetic spectrum monitoring tasks, a fundamental one is detecting the presence of weak communication signal under non-cooperative setting. Various detection methods have been studied in the literature. These detection methods usually construct a proper hypothesis testing metric to distinguish the signal-plus-noise and noise-only signals. Information entropy based metric has recently been studied and shown its superiority over traditional energy based metrics. We find that it can be explained as a realization of the general Kullback-Leibler divergence. Inspired by that, in this paper, we propose to detect the signal using a statistic divergence called β-divergence. Different choices of β will lead to several well known divergences, e.g., Itakura-Saito divergence and Euclidean distance. We study its detection performance for non-cooperative signal sources. Both simulation and real experiments demonstrate that β-divergence can improve detection performance for various modulation types of communication signals.