Otomatik Modülasyon Tanıma Algoritmalarının Geliştirilmesi ve Uygulanması

Abstract

Recognizing features of a signal such as phase, amplitute or modulation type without being noticed by transmitter is called automatic modulation recognition and it is independent from any operator.There are two basic methods for automatic modulation recognition. First one is likelihood based methods second one is feature based methods.During this project, feature based methods are used because feature based methods are more proper for calculations.In this thesis, analog and digital modulation signals are considered.Besides using the higher order cumulants;keyfeature extractions are done by using the characteristics of signals such as instant phase, amplitute and frequency.The feature sets are acquired from different noisy systems,classified by using several classifiers. These classifiers are support vector machines which use keyfeatures,decision trees which use thresholds are obtained from features and hybrid classifier which is combination of decision tree and support vector machines.