EM-based estimators for ARMA parameters of multipath fading channels

Many wireless receiver algorithms including the channel estimators are often based on the assumption that the communication channel is a complex Gaussian process with known mean and autocovariance function. However, the statistics of the fading channel is usually unknown and, therefore, have to be somehow estimated from the received signal. We will show how the frequency selective fading channel can be described by the means of autoregressive moving average (ARMA) parametric models and how the model parameters can be efficiently estimated via the expectation maximization (EM) algorithm and its extensions. Moreover, we will derive computationally attractive recursive estimators for the ARMA channel parameters.