Channel Estimation For MIMO-OFDM Broadband Communication Systems

Orthogonal frequency division multiplexing (OFDM) is an enabling technology for high speed data transmission due to its robustness against frequency-selective fading channels, high spectral efficiency, and simple implementation. Multiple input multiple output (MIMO) greatly increases the system throughput without additional bandwidth and power requirements. Combining the merits of the two technologies, MIMO-OFDM has been accepted as a potential candidate for the physical layer architecture of the fourth generation (4G) systems. However, in fourth generation systems which require high data rate, high quality and high mobility data transmission, accurate channel estimation is an indispensible part. Hence, in this thesis, we mainly focus our research on the topic of channel estimation in MIMO-OFDM systems and represent corresponding solutions in several channel environments. The main contents are as follows:In quasi-static channel, CIR is approximately invariant in one OFDM symbol. A channel estimation method based on delay-subspace tracking is proposed. Firstly, A least square (LS) channel estimation is performed with the aid of optimal pilot sequence of MIMO-OFDM; secondly, the channel impulse response is divided into two parts, and the delay-subspace is tracked using fast data projection method; finally, the desired channel estimation is obtained by projecting CIR onto the delay-subspace. Channel estimation by tracking the delay-subspace greatly improves the estimation quality of LS's. What is more, if we also track the fast changing amplitude of CIR employing adaptive algorithms such as LMS, we will further improve the estimation accuracy.Under certain circumstances such as no pilots available or to save bandwidth, blind channel estimation will be considered. In this paper, we study the principle and conditions of blind channel identification in subspace method, and extend its use to MIMO-OFDM system in static channel. Due to the high dimension of autocovariance matrix of receiving signal, it is difficult to implement blind channel estimation by direct SVD. Therefore, we propose tracking the noise subspace adaptively so as to reduce computational complexity. FDPM, an algorithm with linear complexity, is proved to be stable and robust in noise subspace tracking.In order to reduce channel parameters to be estimated in time and frequency doubly selective channel, we resort to basis expansion model for representing time varying channel and put forward a LMMSE channel estimation based on BEM. To eliminate ICI's adverse effect on OFDM systems in fast varying channel, a block turbo MMSE equalizer is proposed. By exchanging soft information iteratively between the soft output MMSE equalizer and the MAP decoder, the block turbo MMSE equalizer dramatically enhances the equalization performance. Further, we incorporate channel estimation in the iterative equalization and decoding loop and make soft data originating from the turbo equalizer and decoder additional pilots thereby effectively increase channel estimation precision and BER performance.