Radar Emitter Signals Recognition and Classification with Feedforward Networks

Abstract

A possible application of neural networks for timely and reliable recognition of radar signal emitters is investigated. In particular, a large data set of intercepted generic radar signal samples is used for investigating and evaluating several neural network topologies, training parameters, input and output coding and machine learning facilitating data transformations. Three case studies are discussed, where in the first two the radar signals are classified in two broad classes -with civil or military application, based on patterns in their pulse train characteristics and in the third one trained to distinguish between several more specific radar functions. Very competitive results of about 82%, 84% and 67% are achieved on the testing data sets.

Optimization of neural network architecture for classification of radar jamming FM signals

“…Radar jamming signal detection and classification against a background of interference is considered a general radar problem. For military purposes the 'general radar problem' includes searching, interception, localization, analysis and identification of radiated electromagnetic energy which is commonly known as radar electronic support measures [3]. Motivation for this research arises from aiding the movement of countermeasures against unfriendly radar jamming signals by narrowing our focus to classifying their radiated electromagnetic energy.…”

“…Neural networks and their ability to perform radar signal classifications has been the catalyst for extensive investigation [3,[5][6][7][8][9]. This has led to an increasing effort in the research to optimize NN architectures and topologies to ensure acceptable and high efficiency classification results [10,11].…”

“…Petrov, Jordanov, and Roe [3] attempted to increase classification accuracies between civilian and military signals by examining two NN architectures with different neural topologies in the output layer: 1 output neuron in the first case and 2 output neurons in the second. The rest of the architectural characteristics were kept identical.…”

“…The preprocessing of the available data is of great importance for the machine learning stage and usually can significantly affect the overall success or failure of the application of a given classification problem [3]. A transformation may be necessary to eliminate insignificant variations and superfluous details while at the same time accentuating the pertinent information [7].…”

Lyapunov stability criterion based neural inverse tracking for unknown dynamic plants

“…In the past few decades, artificial neural networks have emerged as powerful computational tool and have been successfully employed in various real world applications that include business and economics [1][2][3]; forecasting [4,5]; sports [6,7]; medicine [8,9]; and engineering [10][11][12]. The massively parallel and redundant architecture of neural networks lends themselves the important characteristic of function approximation to an arbitrary degree of accuracy [13].…”