Abstract: A major challenge in broadband MIMO-OFDM system is how to obtain the Channel State Information (CSI) accurately and promptly for coherent detection of information symbols and channel tracking. This thesis formulates the channel estimation problems for MIMO-OFDM systems and proposes a pilot-tone based estimation algorithms. The complex equivalent baseband MIMO-OFDM signal model is derived using matrix representation. The procedure is summarized as: First, pilot-tone arrangement is presented in both mathematical and simulation bases, from which a proposed scattered pilot-tone pattern arrangement is combined with Least Square (LS) estimation algorithm. The designed pilot-tone arrangement uses .. pilot-tones (.. represents the maximum channel impulse response length) in each OFDM symbol. The results are obtained for frequency selective fast fading channels. The channel is assumed to be quasi-static Rayleigh fading throughout the simulation, i.e., the channel does not change its state in one OFDM symbol. Secondly, this thesis proposes adaptive weights algorithm for MIMO-OFDM detection based on scattered pilot-tone arrangement and LS estimation in fast fading and frequency selective Rayleigh fading channels. Mathematical and simulation analysis are given for three proposed scattered techniques: Adaptive weights LS, adaptive weights Minimum Mean Square Error (MMSE), and adaptive weights Vertical Bell Laboratories Layered Space Time (V-BLAST). Simulations are carried out for different antenna sets (2×2, 3×3 and 4×4). The results obtained show that the proposed techniques have the ability to track CSI from one OFDM to another when compared to the conventional preamble-based system that uses one or two OFDM symbols as training at the beginning of a frame. For example, a 4×4 MIMO-OFDM system having an FFT size of 512 and a channel impulse response length of 16, the algorithm achieves less than 10-3 bit error rate at signal-to-noise ratio of 30 dB. While the preamble based system fails to track the channel because of lack of availability of channel information. Performance comparison among the proposed techniques and the corresponding results are compared with those related to the conventional technique. Thirdly, this thesis proposes modification to the well known IEEE802.11a standard in the physical layer. This standard adopts OFDM as the main modulation scheme and uses a preamble of 4 OFDM symbols at the beginning of the frame to estimate a channel in the frequency domain. The proposed modification is to use multiple antennas at both sides and to estimate the channel using time domain. A sample result is obtained for 3×3 MIMO-OFDM IEEE802.11a modified system for different mapping techniques. The results obtained show that enhancement in performance is achieved in both data rate and diversity order. As an example, BER = 10-5 at SNR = 20 dB for QPSK in slow fading channel and BER = 5×10-5 at SNR = 30 dB for 64QAM in slow fading channel.