Orthogonal frequency division multiplexing (OFDM) has been received much attention for high rate transmission wireless communication systems such as future-generation mobile communication systems, wireless Internet access, and wireless multimedia due to its bandwidth efficiency and robustness against the frequency-selective channels. In addition, OFDM is a promising digital modulation scheme to simplify the equalization in the frequency-selective channels. Recently, by using multiple antennas at the transmitter and the receiver, we can exploit the diversity gain and solve the high link budget problem required at the high transmission system by combining the OFDM with the MIMO system. To maximize the performance advantage of multiple-input multiple-output OFDM systems, a coherent signal detection requires reliable channel impulse response between transmit and receive antennas. However, since the channels are time-varying and frequency-selective in broadband wireless communication systems, the channel estimation is known to be one of the difficult tasks in MIMO systems. The channel estimation can be performed by either inserting pilot symbols into all of the subcarriers of OFDM symbols with a specific period or inserting pilot symbols into each OFDM symbols. However, since the pilot-based channel estimation requires many pilot symbols and decreases the transmission bandwidth, it is not bandwidth efficient in terms of data transmission. In order to overcome the disadvantage of the pilot-based channel estimation, subspace-based (semi-)blind channel estimation methods have been studied extensively. The subspace-based (semi-)blind channel estimation methods have a simple structure and achieve good performance, but usually require large number of data blocks because the many existing (semi-)blind channel estimation methods are statistical in nature (e.g., second-order statistics).
In this thesis, we propose the statistical channel estimation for MIMO-OFDM systems by...