Abstract

In this paper, automatic modulation recognition of mixed multiple source signals is discussed. At first, the algorithm of equivariant adaptive source separation (EASI) is employed to separate signals from their mixed waveforms. Four features of five modulated signals are extracted and then two classifiers, decision tree and neutral network are used to complete modulation classification. The effects of symbol shaping on features extraction and validation of source separation are also investigated. Simulations show that the average probability of correct recognition of the classifiers is very depended on the performance of source separation. When SNR (Signal to Noise Ratio) is larger than 15 dB and the number of mixed source signals is less than 4, the average probability of correct recognition is above 0.6 for decision tree classifier and 0.63 for neutral network classifier. Simulations and discussions about automatic modulation recognition for source signals surfed Rayleigh flat fading are also presented.