Modulation classification using ARBF networks

A simple and robust method based on statistical pattern recognition theory to approach modulation type classification is proposed. The features being used for classifying digital signaling formats are the fourth-order and sixth-order cumulants of the received signal. An adaptive radial-basis function networks (ARBF) which tandem combine a single-layer RBF and a single-layer linear-basis function (LBF) networks is constructed as the classifier. Examples of classifying four modulation types - 4 ASK, 2 ASK/2 PSK, 4 PSK and 16 QAM - are given. T´he result of computer simulation has proved that this method can process well in a wide range of SNR and has a preferable generalization.