Basis Expansion Model-based Doubly Selective Channel Estimation And Detection Techniques For Multi-carrier Communication Systems

Orthogonal frequency division multiplexing(OFDM) system has been considered as a main wireless transmission technology due to its high spectrum efficiency. In order to recover the transmitted symbols in coherent demodulation, channel state information plays an important role for OFDM systems. However, for the high-speed mobile environment, OFDM systems meet doubly selective fading channels, leading to a challenge issue of channel estimation. Different from traditional channel models, basis expansion model(BEM) can use the equivalent description of transform domain for transmission channel to efficiently reduce the dimension of unknown channel parameters. Thus, BEM has drawn a lot of research interests in the literature. In this dissertation, the researches on channel estimation by following BEM are discussed as follows.Firstly, channel estimation for single-input single-output orthogonal frequency division multiplexing(SISO-OFDM) systems is studied for doubly selective fading channels. The asymptotical relationship between high-order Wiener filters(WF) and the linear interpolation is derived in detail. According to the correlation of wireless channels and the approximate nature of Bessel functions, a low-complexity three-order WF is proposed in this dissertation. In addition, both theoretical analysis and simulation results prove the effectiveness of the proposed low-complexity channel estimation method. A time-domain WF based on BEM is proposed, which performs better in high-speed environment by more accurately modeling inter-carrier interference(ICI) caused by doubly selective channels.Secondly, joint channel estimation and data detection for centralized MIMO-OFDM systems is studied. Using the expectation maximization(EM) algorithm, an approximate maximum likelihood(ML) iterative updating method of joint channel estimation and data detection for multiple antennas is proposed, which allows the unknown background noise variance. Theoretical analysis and simulation results prove that the proposed method approaches the ML lower-bound. Meanwhile, in order to reduce the computational complexity, a sub-optimal method which independently updates the channel and data information of each antenna is proposed, by expressing the received signal in the form of superimposing each of the transmitted signals. Simulation results show that, compared to the proposed joint estimation and detection method, the proposed sub-optimal method can achieve a better compromise between performance and complexity.Thirdly, joint channel estimation and data detection for distributed MIMO-OFDM systems under complex environment is studied. Except for the unknown background noise, the interference caused by spectrum sharing and the phase noise caused by instability of oscillators are also considered in the dissertation. Using the empirical priori information, according to the maximum a posterior probability(MAP) criterion, a variational Bayesian inference(VBI)-based method that jointly optimizes channel parameters and data symbols is proposed. Simulation results show that the proposed method approaches the performance lower-bound, which assumes perfect channel state information, no phase noise, and known positions and powers of interference plus additive white Gaussian noise.Finally, joint channel estimation and data detection for OFDM-based cooperative two-way relay network(TWRN) in non-reciprocal doubly selective fading channels is studied. In this dissertation, an improved initialization method using minimum mean square error(MMSE) criterion is proposed by assuming the background noise variance is known. An approximate MAP detection algorithm is proposed by using the empirical prior information and VBI. Simulation results show that the performance of the proposed VBI method approaches that of the ideal case, which assumes perfect knowledge of channel state information.In summary, by using the BEM which effectively represents doubly selective fading channels, the issues of channel estimation for OFDM-based communication systems are investigated in this dissertation. Various channel estimation methods are studied under the Bayesian framework. The application of the prior information to complex environments where background noise, interference and phase noise are unknown is demonstrated.