Graph-based joint channel estimation and data detection for large-scale multiuser MIMO-OFDM systems /

Thesis Type: Postgraduate

Institution Of The Thesis: Orta Doğu Teknik Üniversitesi, Faculty of Engineering, Department of Electrical and Electronics Engineering, Turkey

Approval Date: 2015




In this thesis, a graph-based soft iterative receiver for large-scale multiuser MIMO-OFDM systems is proposed that performs joint channel estimation and data detection over time-varying frequency selective channel. In an uplink scenario, factor graph structures for the transmitter of users and the receiver of base-station are presented, which provide Gaussian message passing between nodes. Instead of LLR, reliability information of symbols are used to decrease complexity of the proposed algorithm. Training symbols, known at the receiver, are utilized to get channel state information at the initialization. Also a new training structure is proposed which enables channel estimation and data detection for numerous users. Soft channel estimation process is introduced which utilizes correlation information between channel coefficients. Transfer nodes bring reliability information of channel coefficients between coefficient nodes to converge actual value. Message passing schedule is rearranged to enhance performance of the graph based soft iterative receiver. Extrinsic information exchange is applied between nodes of the repeated symbols. Soft information of the channel coefficients and symbols are jointly refined in each iteration. The BER performance analysis of graph based soft iterative receiver is investigated by comparing non-iterative ML and MRC. Simulation results show that the proposed algorithm with channel knowledge has a similar performance with MRC and outperforms non-iterative ML. Performance of GSIR with different training symbol spacing, number of users, number of receive antennas, code rates and constellations are compared to provide an overview of the proposed algorithm. Also channel estimation performance of GSIR is analyzed by comparing with perfect channel knowledge case. A LDPC decoder is used in combination with GSIR to increase total performance.