In this dissertation, the modeling of multiple-input multiple-output (MIMO) time-varying flat fading channels is studied first, then several second-order statistics of eigen-channels and the instantaneous mutual information (IMI) in such channels are explored. Finally, efficient estimation of MIMO frequency-selective fading channels is investigated.

Accurate characterization of MIMO fading channels is an important prerequisite for the design of multi-antenna wireless communication systems. Here, a new statistical model for time-varying MIMO flat fading channels is proposed, and a closed-form expression for the spatio-temporal cross-correlation function between any two subchannels is derived. This new analytical correlation expression includes key physical parameters of interest such as the mean angle-of-departure and mean angle-of-arrival, the associated angle spreads, Doppler spread, and others, in a compact form. The utility of the proposed model is revealed by a comparison with the collected MIMO data in terms of the spatio-temporal correlations, level crossing rate, average fade duration, and the cumulative distribution of IMI.

Although they provides useful information for rate control in multiuser communication systems, the second-order statistics of eigen-channels and IMI in time-varying fading channels have not been studied so far, to the best of our knowledge. In this dissertation, first the autocorrelation function, correlation coefficient, level crossing rate, and average outage duration of IMI are investigated in maximal-ratio-combiner (MRC) MIMO systems. Then the results are extended to general MIMO systems with the aid of random matrix theory. Moreover, the statistical property of eigen-channels is also studied in general MIMO systems. Monte Carlo simulations are provided to verify the accuracy of the analytical results. The results shed more light on the dynamic behavior IMI in mobile fading channels.

Optimal training sequences design is an important issue in the successful deployment of MIMO. However, optimal design with low implementation complexity is a challenging task for MIMO frequency-selective channels. In the last part of this dissertation, optimal training sequence design for estimating MIMO intersymbol interference (ISI) channels is addressed. In addition, several novel low-complexity channel estimators are proposed using uncorrelated Golay complementary sets of sequences. The theoretical analysis and simulations show that when the additive noise is Gaussian, the proposed best linear unbiased estimator (BLUE) achieves the minimum possible classical Cramér-Rao lower bound (CRLB) if the channel coefficients are regarded as unknown deterministics. On the other hand, the proposed linear minimum mean square error (LMMSE) estimator attains the minimum possible Bayesian CRLB, when the underlying channel coefficients are Gaussian and independent of the additive Gaussian noise. The proposed channel estimators not only achieve the best estimation performance, but also can be implemented with low complexity, via DSP or ASIC/FPGA. This is possible due to the special structures intrinsic to uncorrelated Golay complementary sets of sequences, making the proposed channel estimators ready to use in practical MIMO systems.

Committee Members:

Dr. Ali Abdi (Advisor), Assistant Professor, ECE Dept., NJIT

Dr. Yeheskel Bar-Ness, Distinguished Professor, ECE Dept., NJIT

Dr. Hongya Ge, Associate Professor, ECE Dept., NJIT

Dr. Alexander M. Haimovich, Professor, ECE Dept., NJIT

Dr. Andreas F. Molisch, Professor, Department of Electroscience, Lund University, Lund, Sweden

Senior Principal Technical Staff, Mitsubishi Electric Research Labs, Cambridge, MA