Blind detection in symmetric non-Gaussian noise with unknown PDF using maximum entropy method with moment generating function constraints

The paper provides a blind binary detection approach in an unknown non-Gaussian noise. In our scheme, we use maximum likelihood (ML) detection rule in conjunction with maximum entropy method (MEM) for probability density function (PDF) estimation of the unknown observation noise from the samples of the received data. We constrain MEM on estimated moment generating function (MGF). The estimated PDF based on MEM–MGF is quite close to the true PDF and has a direct applicability for blind implementation. The results indicate that the new nonlinear detector outperforms conventional matched filter, and approaches the performance of the optimal ML detector which assumes the complete knowledge for the noise PDF. Then, we analyze the scheme by probability of error (PePe) calculation and interpret the results.