Abstract
Recognition of multifunction radar (MFR) is an open problem in the field of electronic intelligence. Parameters of MFR pulses are generally agile and difficult to distinguish statistically. A prospective way to realize credible MFR recognition is mining and exploiting more distinguishable high-dimensional patterns buried in pulse groups, which may be designed for implementing infrequently used radar modes such as target tracking. A high-dimensional pattern is defined according to the agile range and switching law of sequential pulse repetitive intervals within a pulse group. This article establishes deep recurrent neural networks (RNN) to discriminate and coarsely cluster different pulse groups hierarchically with respect to their sequential structures. Afterwards, RNN-based classifiers are trained to extract and exploit features within different pulse group clusters. Distinct degrees of confidence are then attached to these classifiers to indicate the discriminabilities of the corresponding pulse group clusters. The pulse group clustering and classifying models are finally cascaded to form an integrated classification model, which mines distinguishable patterns from sequentially arriving pulse groups of the same radar and accumulate them to realize MFR recognition. Simulation results demonstrate the much improved performance of the proposed method over existing counterparts in different scenarios.
Highlights
Multifunction radars (MFR) are widely used in civil and military areas [1]–[4], they usually have multiple modes such as storm surveillance and tracking [5], target searching and tracking [6], [7]
pulse repetitive interval (PRI) sequence clustering models are trained with module “hierarchical clustering of pulse groups” based on a dataset including a large amount of pulse groups, and recurrent neural networks (RNNs) models (Ri) that are able to cluster pulse groups according to their patterns will be obtained
The regularly used searching mode of all radars is performed with equi-PRI pulse groups, which account for a large proportion in the observed data
Summary
Multifunction radars (MFR) are widely used in civil and military areas [1]–[4], they usually have multiple modes such as storm surveillance and tracking [5], target searching and tracking [6], [7]. RNN extracts high-dimensional patterns buried in pulse trains to realize radar classification, and the method achieves satisfactory performances in cases of agile pulse parameters and significant observation noises [18] This method may not work for MFRs. That is because deep learning methods usually have strong statistical tendency, i.e., they incline to learn from and match predominant modes in the given dataset. By considering that frequently and infrequently used modes of MFR may differ largely in discriminability and take much differentiated proportions in noncooperatively observed signals, this article trains a series of RNNs hierarchically to cluster the pulse groups blindly and extract sequential PRI patterns from them to realize MFR recognition.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
More From: IEEE Transactions on Aerospace and Electronic Systems
Disclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.