Blind classification for linear and non-linear modulations based on the fusion of multiple features

Abstract

Blind identification for modulations is an important issue in signal processing and wireless communications. The role of modulation identification is to find out which type of modulations for the signals received. To investigate the classification for both linear and non-linear modulations, the fusion of multiple features is studied in terms of the cumulants, approximate entropy and kurtosis. The features are combined as the input vector of back propagation neural network, which is designed to discriminate multiple modulations. Training and test are verified via simulations finally.