Abstract
This paper addresses the problem of channel estimation in multiple-input and multiple-output orthogonal frequency-division multiplexing (MIMO-OFDM) systems for the case when the underlying multipath channel is approximately sparse in the angle-delay domain. To this end, an algorithm for deterministic selection of pilot tones over which (either random or deterministic) training vectors can be transmitted is proposed. In addition, conditions are derived for the minimum number of pilot tones needed by the proposed algorithm to ensure reliable estimation of the underlying channel using reconstruction methods from the compressed sensing literature. Finally, effectiveness of the proposed algorithm is demonstrated through numerical experiments.