Frequency-domain channel estimation in MIMO-OFDM

In this work, we propose a simple, yet flexible channel estimator for MIMO-OFDM systems. It works in the frequency domain, by interpolating among the estimates done in a small number of subcarriers. Our estimator can be easily used in either acquisition (preamble-based) or tracking (pilot-tones based) mode, and its structure remains the same for any type of training pattern in the two-dimensional time-frequency space. We also propose efficient preambles that allow identification of the MIMO channel with the lowest sacrifice in data rate. The training sequence we advocate consists of a set of OFDM symbols endowed with orthogonality properties, whose duration is much lower than the number of OFDM subcarriers, and which is robust against frequency misalignments. The feasibility of our approach is substantiated by computer simulation results obtained for IEEE 802.16 broadband fixed wireless channel models.