Multislot estimation of frequency-selective fast-varying channels

In mobile communications, the movement of terminals renders the multipath channel time varying. Even though the faded amplitudes are fast varying, the delays can be considered as stationary on a large temporal scale. In this paper, we propose a new subspace-based method that estimates the channel response from multiple slots by capitalizing on these different varying rates without explicitly computing the delays of the multipath. The temporal subspace is obtained from multiple single-slot training-based estimates of the (single-user or multiuser) channel response. Provided that the number of slots is large enough, the time basis can be calculated with any accuracy. As a consequence, the mean-square error on the channel response depends only on the number of fast-varying parameters that have to be estimated in a slot-by-slot fashion. Performance analysis and simulations confirm the expected benefits of the multislot approach in improving the efficiency of systems with short training sequences.