Extended Least Squares Identification of Doubly Spread Mobile Communication Channels

This paper presents an on-line coupled filter extension to the method of Tsatsanis et al. [1] for identification of randomly time-varying linear channels. The channel is modelled as a tapped-delay line filter, with tap coefficients described by an unknown auto-regressive (AR) process. We propose an augmented-state Kalman filter for tracking the channel taps and estimating the mean channel response. The recursive algorithm of [1] for estimating the channel covariance and AR parameters on-line is modified to incorporate the mean channel estimate. This results in a coupled estimator structure in which the AR parameters and mean channel response are estimated in parallel. We also note the relationship of the covariance estimation technique to maximum likelihood estimation. Simulation studies demonstrate the performance of the proposed estimator.