Modulation classification based on a maximum-likelihood receiver in the cyclic-HOS domain

A multiple hypothesis QAM modulation classification task is adressed in this paper. The classifier is designed in the framework of decision theory. A characteristic feature is extracted from the signal, and is compared to the possible theoretical features in the maximum-likelihood sense. This feature is composed of a combination between fourth-order and squared second-order cyclic temporal cumulante. The combination between cumulante of different orders is intended to bypass the incertainty about the power of the signal of interest. As an application, we present simulated performance in the context of 4-QAM vs 16-QAM vs 64-QAM classification.