Digital modulations automatic classification using the combination of several features extracted from time and time-frequence domain

This paper presents a method for the automatic classification of digital modulations including ASK,PSK,QAM,FSK,MSK,LFM,OFDM, without a priori knowledge of the signal parameters by a large set of features which are extracted from ambiguity function (AF), Choi-William distribution (CWD) and time domain. In this method the whole features are extracted and pruned by discarding redundant features using principal component analysis (PCA) technique, then inputted to the SVM-based classifier for decision-making. To evaluate the method performance, some simulations have been carried out in different conditions of signal noise ratio and the main parameter. To assess the advantages, comparison with other classification methods has been given. The final results verified the feasibility and effectiveness of the algorithm in this paper,which also proved the good robustness of the method in condition of low signal noise ratio.