Complex Convolutional Neural Network for Signal Representation and Its Application to Radar Emitter Recognition

Abstract

Signal representation and identification in wireless communication has recently aroused substantial concern due to the remarkable evolution of data-driven techniques. Consider there are vital complementary information between inphase (I) and quadrature (Q) components, and I and Q have respective emphasis on the signal. Therefore compact signal representation requires their joint interaction. However, it is a pity that most methods ignore their implicit relevance. This letter considers the codify of tacit knowledge between I/Q pairs. First, a complex-valued convolutional unit is proposed to mine individual information and to explore their complementation. Second, a complex convolutional neural network (CCNN) is built to achieve radar emitter recognition end-to-end. Experimental results demonstrate CCNN’s superiority on measured radar signals in comparison to other methods from literature.