USING NEURAL NETWORK IN RADAR EMITTER'S IDENTIFICATION

Document Type : Original Article

Authors

Abstract

A region of the microwave band can contain many radar signals from different emitters. There are few techniques for identification of emitters on the base of measuring the parameters of their received signals. However, these techniques are utilized OFF-line. In modern electronic warfare, the ON-line techniques in the processing and decision are required. It is assumed that the electronic support measurement (ESM) receiving system is used to measure radar parameters. These parameters are : the carrier frequency (F), the pulse width (PW), the antenna scan rate (RPM), the angle of arrival (A.OA) and the time of intercept (TOI) for each pulse. Data processing in ESM contains two basic functions: deinterleaving (sorting or classification) and emitter identification. ESM systemn applies some algorithms to classify the radar emitters and the resulting radar parameters F, PW, PRF, RPM are called the emitter descriptor vector. They are used in emitter identification by comparison with an existing library. This paper suggests a novel technique to identify radar signals using neural network on simulated radar parameters. A fully connected feed forward network (back propagation) is designed and implemented. The designed network consists of three layers. Th(z input layer contains four processing elements that correspond to the four measured radar parameters resulting from the classification process. The hidden !Layer consists of five processing elements. The output layer consists of ten processing elements that correspond to the number of emitters in the library. The proposed network is tested by several examples to verify the design concept.