Modulation recognition using artificial neural networks

Abstract

This paper presents artificial neural networks (ANNs) for the recognition of either analogue or digital modulation types. Computer simulations of different types of band-limited, modulated signals corrupted by band-limited Gaussian noise sequence have been carried out to measure the performance of the ANN approach. The threshold SNR for the recognition of either analogue or digitally modulated signals with average success rate ≥ 98% is found to be about 10 dB. Comparisons of results from the ANN approaches and the decision-tree methods are presented. © 1997 Elsevier Science B.V.