A covariance-convolution framework for radar intra-pulse modulation classification

In this paper, a signal processing algorithm for modulation classification of pulse-compressed threat radar signals by an airborne, electronic warfare (EW) digital receiver (DRx) is presented. Modern radars employ pulse compression techniques like frequency modulation (FM), phasemodulation (PM), frequency hopping (FH), etc. to achieve specific operational objectives. State-of-the-art technologies and systems enable broadband signal design, intra-pulse techniques and joint-domain analysis in radars. To counter modern radars, EW systems need to have real-time adaptive capabilities, in multiple domains, over the vast multi-parametric space of radar operations. Current work focusses on the challenges associated with intra-pulse modulation classification, pivotal for scenario cognition. Novelty of the proposed method is its applicability for analysis of arbitrary intrapulse techniques, by treating pulse compression as continuous time (CT) or discrete time (DT) process and classification based on the underlying signal structure. The method is a two-level classifier designed for four classes of signals viz. Unmodulated (Unmod), FM, PM and FH. In the first level, classification is done between CT and DT modulations. By design, Unmod & FM are CT modulations, while PM & FH are DT modulations. In the second level, classification is done between Unmod & FM and between PM & FH. The method utilizes covariance, convolution and stationarity properties of signals of interest. Classification performance in multi-parametric space of each modulation type, is quantified in terms of confusion probabilities over SNR. Efficacy of proposed method has been ascertained in the actual EW DRx, against signals generated using radar signal design library & advanced signal generators.