A novel algorithm for automatic constellation classification of PSK and QAM signals and a RBF-based identification

The authors present a novel algorithm to automatically classify M-ary QAM and M-ary PSK signals in the presence of additive white Gaussian noise (AWGN) into various constellation types and then identify them using radial basis function (RBF) neural networks. The signals are first passed through a two-layer RBF network processing structure, wherein the signals are trained to be distinguished into the various constellation points using a modified recursive-training algorithm. The proposed algorithm being flexible can be easily expanded to identify all the M-ary QAM and PSK constellation types. The performance of the algorithm is evaluated using simulations in MATLAB to illustrate the effectiveness.