Complex-valued neural polynomial adaptive equalizer for digital satellite channels

An adaptive equalizer using a complex-valued neural polynomial structure (CNPAE) is proposed for digital satellite channels in this paper. Since CNPAE consists of a single layer neural network and nonlinear processor, its structure is very simple. Moreover, it has fewer parameters than traditional methods (Multilayer neural networks, recurrent neural networks, Volterra filter and etc.). To evaluate the performance of the CNPAE, it is applied to solve nonlinear equalization problem of a digital satellite communication channel for M-PSK, which has server nonlinearity with memory. Simulation results illustrate the effectiveness of the CNPAE.