Subspace-Based Blind Channel Estimation and Tracking for MIMO-OFDM Systems

Multiple-input multiple-output (MIMO) orthogonal frequency division multiplexing (OFDM) is now widely considered as a favored technology for emerging and future generation wireless systems. MIMO-OFDM aims to achieve increased channel capacity limit by exploiting the use of multiple antennas in combination with multi-carrier orthogonal modulation. While the possibility of achieving this limit is bestowed on the invention of capacity-achieving coding and decoding techniques, in reality, this prospect relies heavily on the existence and use of advanced channel estimation techniques. To facilitate fast and reliable channel estimation in MIMO-OFDM systems, pilot symbol insertion is usually considered; however, the channel capacity is greatly reduced by inserting those pilot symbols. Therefore, employing fast-converging and reliable blind channel estimation for MIMO-OFDM seems to be an attractive solution for future wireless systems.;In this thesis, we propose a new subspace-based blind channel estimator that requires only a comparably short time averaging period. We consider the design of such an estimator directly in the frequency domain, as opposed to the majority of existing designs in which estimators are developed in the time domain. Our first contribution is to propose and investigate a novel subspace-based estimator with reduced time averaging, by exploiting the frequency correlation among adjacent subcarriers, residing within the coherence bandwidth of the broadband channels in typical MIMO-OFDM scenarios. To reduce the high computational complexity incurred by the eigenvalue decomposition and the associated ambiguity matrix, our second contribution is to develop an improved, adaptive version of the estimator for enhancing its capability under MIMO time-varying conditions. This is achieved by employing a modified form of the orthogonal iteration for efficient subspace tracking along with a precoding technique that allows a reduction in the size of ambiguity matrix. Numerical experiments demonstrate that the proposed techniques can indeed outperform several benchmark estimators in various practical scenarios.;To this end, blind channel estimation based on second order statistics (SOS), instead of higher order statistics (HOS), has been widely considered as a suitable candidate. Amid SOS-based blind approaches, subspace-based estimation is attractive since reliable estimates can often be obtained in a simple form by optimizing a quadratic cost function. Nonetheless, the performance of the subspace-based blind channel estimators may still be seriously degraded under time-varying conditions. This problem can generally make overall performance unsatisfactory, especially in MIMO-OFDM systems whose number of subcarriers is large. In order to overcome this limitation and successfully employ subspace-based channel estimation in MIMO-OFDM systems, it is essential to minimize the required length of the underlying time averaging period.