C25. Digital signal classification by compressed cyclostationary features

Abstract

Different classifiers have been adapted for classification of digital signals in low SNR environments in many researches. For efficient performance of signals classifiers and features detectors in real time, limited number of features are required. In this paper we introduce a method to compress the cyclostationary features of digital signals using Discrete Wavelet Transform (DWT). The target is to reach low percentage of classification error with reducing the number of features required. Using the proposed technique, different types of digital signals as BPSK and QPSK signals in three different cognitive radio scenarios have been considered. Simulation results show that the proposed technique can achieve classification accuracy up to 97% with a reduction percentage of 87.5% in the utilized features.