Neural Networks for Signal Intelligence
Jithin Jagannath, Nicholas Polosky, Anu Jagannath, Francesco Restuccia, Tommaso Melodia · 2019
This chapter provides readers with a comprehensive account of how machine learning (ML) techniques, specifically artificial neural networks, have been applied to solve some of the key problems related to gathering signal intelligence. To accomplish this, it begins by presenting an overview of artificial neural networks. Then, the chapter discusses the influence of ML on the physical layer in the context of signal intelligence. ML techniques for signal intelligence typically manifest themselves as solutions to discriminative tasks. That is, many applications focus on multi-class or binary classification tasks. Perhaps the most prevalent signal intelligence task solved using ML techniques is that of automatic modulation classification. In short, this task involves determining what scheme was used to modulate the transmitted signal given the raw signal observed at the receiver. Other signal intelligence tasks that employ ML solutions include wireless interference classification. The chapter also discusses directions taken by the community towards hardware implementation.