Classification of Radar and Communications Signals Using Wideband Autonomous Cognitive Radios

Spectrum awareness is one of the most challenging problems in wideband autonomous cognitive radio (WACR) design. Detection and classification of low-SNR signals is important for proper WACR functionality as it enables the radio to adapt to the user needs and surrounding RF environment. In this context, identification of radar and communications signals is critical in various applications, especially, electronic warfare. This paper introduces a classification framework for radar and communications signals based on their cyclostationary features, specifically, the cyclic profile. Two classification algorithms are used: an artificial neural network (ANN) and a convolutional neural network (CNN). The simulation results show that cyclic profile is a good candidate compared with other features to distinguish between radar and communications signals even at very low SNRs. Furthermore, from a complexity perspective, the ANN is shown to be more effective than the CNN in the proposed classification framework.