Signal classification based on spectral redundancy and neural network ensembles

In the last couple of decades, the introduction of new wireless applications and services, which have to coexist with already deployed ones, is creating problems in the allocation of the unlicensed spectrum. In order to overcome such a problem, by exploiting efficiently the spectral resources, dynamic spectrum access has been proposed. In this context, cognitive radio represents one of the most promising technologies which allows an efficient use of the radio resource by collecting, processing and exploiting information regarding the spectrum utilization in a monitored area. To this end, in this paper the problem of classifying similar signals characterized by different spectral redundancies is addressed by using a neural network ensemble. A set of simulations have been carried out to prove the effectiveness of the considered algorithms and numerical results are reported.