Channel Model Identification for Godard-Kalman Modeling of Time-varying Communication Systems

This paper presents a Godard-Kalman approach with adaptive model parameters identification for model-based adaptive filtering over time-varying communication channels. The presented approach enables model-based adaptive channel equalization without prior channel estimation. An adaptive identification of autoregressive (AR) model coefficients is performed to overcome the issue of determining model coefficients which capture the dynamics of unknown time-varying channels. Experimental MSE performance of the adaptive algorithms are simulated in a multi-user environment, assuming a vector AR(1) model for the optimal filter weighs. Superior performance of the Godard-Kalman algorithm with adaptive model identification is demonstrated, comparing to the same algorithm with fixed model coefficients and to standard observation-only-based LMS and RLS adaptive algorithms.