Correlation based channel estimation algorithm for autoregressive channel models with increased robustness

An estimation algorithm with improved robustness is presented for the class of linear channels that can be modelled by an autoregressive (AR) model. It combines features of two previously developed concepts: the finite correlation LMS (FCLS) and dynamic based deconvolution model (DBD). The new correlation DBD (CDBD) exploits the projection properties of the FCLS, leading to enhanced performance, as well as the FCLS correlation properties, leading to noise robustness, with the high adaptation performance of the DBD. Results are shown for uncoded and chaotic coded signals, where chaotic coding is achieved using the logistic map. The later is easy to implement in real time applications. The sensitivity of the CDBD to model uncertainties is addressed and it is shown that these do not influence its performance significantly.