Adaptive exploitation of cyclostationarity features for detecting modulated signals

In this paper, a detector is presented that can be trained to recognise the dominant features in the cyclic spectrum of a signal, and hence distinguish between scenarios when the signal is present or absent. A detector that can adapt its operation, without knowing in advance what spectral features to expect, is attractive for applications such as cognitive radio, as equipped terminals can learn about their radio environment whilst operating in the field. The performance of this detector is compared to that of an energy detector via simulation.