9650

8894

كارشناسي ارشد

مخابرات﴿سيستم﴾

اصفهان: دانشگاه صنعتي اصفهان، دانشكده برق و كامپيوتر

1393

چهارده،90ص.: مصور،جدول،نمودار

ص.ع.به فارسي و انگليسي

محمدرضا احمدزاده

93/12/19

عبدالرضا ميرزايي

مهندسي برق و كامپيوتر

ID8894

Recognition and Classification of Radar Emitter with Discriminative Features Seyed Moslem Shokrolahi m shokrolahi@ec iut ac ir Date of Submission 2015 01 11 Department of Electrical and Computer Engineering Isfahan University of Technology Isfahan 84156 83111 Iran Degree M Sc Language FarsiM R Ahmadzadeh Assist Prof supervisor Ahmadzadeh@cc iut ac irAbstract The radar emitter recognition seems to be one of the most important tasks in electronic supportmeasures ESM and electronic intelligence ELINT systems The main function of electronic sup port measures system is threat detection and area surveillance to determine the identity of surroundingemitters radars A typical intercept receiver system in ESM must be able to intercept signals fromantenna or antenna array and extracts the basic parameters features into structures called pulse de scriptor words PDWs These basic parameters usually contain the values for radio frequency RF pulse width PW direction of arrival DOA pulse repetition interval PRI time of arrival TOA andpulse amplitude PA After determining these parameters it is possible to identify the specific emit ters The radar emitter recognition consists of two major goals First goal is determining the numberof emitters present Second objective is classifying or recognizing emitters according to a library ofknown emitter characteristics In this thesis we use real radar signal data which are recorded from aspecific area to develop a radar emitter recognition system In this thesis we propose a three layer model for achieving specific emitter identification A newfeature subset selection method via multi objective particle swarm optimization and gap statistic clus tering criterion is proposed in order to determine the number of emitters and eliminate superfluousfeatures Then we propose a new hybrid clustering algorithm by combining Rough k means and impe rialist competitive algorithm Finally radar data are classified by using artificial neural network ANN To improve the performance of the classification layer we use hybrid training algorithm We combineback propagation algorithm with genetic particle swarm optimization and imperialist competitive al gorithms to avoid getting trapped in local minima Experimental results show that our proposed modelhas better performance in comparison with other methods Keywords Radar pulse descriptor words feature selection Rough k means artificial neural network clustering

محمدرضا احمدزاده

عبدالرضا ميرزايي