Detection of stochastic signals in narrowband non-Gaussian noise

A general receiver is developed for the optimum detection of stochastic signals in narrowband non-Gaussian noise. Exploiting the Middleton Class A noise model, we obtain a generalization of the estimator-correlator structure: the likelihood ratio processor involves a series of estimator-correlators which then get exponentiated, weighted and summed. It appears that for most situations of interest the series involves just a few terms; as indicated by preliminary examples, three or four terms generally suffice to closely approximate the optimum receiver.