Research On Channel Estimation And Signal Detection For Future 5G Systems

To cater the needs of future mobile communication applications,based on the key technologies of the fifth generation(5G)system,including massive multipleinput multiple-output(MIMO)technology,non-orthogonal multiple access(NOMA)technology and index modulation(IM)technology,this thesis investigates signal detection and channel estimation algorithms based on message passing(MP)method.Firstly,multiuser signal detection in massive MIMO systems is investigated.Based on MP,five multiuser iterative detection schemes are proposed for three different massive MIMO systems.In the case of the base station(BS)having the perfect channel state information(CSI),expectation propagation(EP)based detection scheme 1 and linear minimum mean square error(LMMSE)based detection scheme 2 are proposed for multiuser signal detection.Simulation results show that the detection performance of the two proposed detection schemes 1 and 2 is very similar.And the both achieve significant detection performance gain over the conventional LMMSE detector.This is because the proposed detection schemes 1and 2 can make full use of the priori information of the users signal.Moreover,in terms of the computational complexity,the proposed detection scheme 1 is one order of magnitude lower than that of the proposed detection scheme 2 and the conventional LMMSE detector.In the case of estimating channels in the BS,an EP based detection scheme 3 is proposed for multiuser signal detection with estimated channels.Simulation results show that the performance of detection scheme3 is better than that of detection scheme 1.The main reason for this is that the detection scheme 3 considers the impact of channel estimation error,while the detection scheme 1 does not,which naturally results in the performance loss of detection scheme 1.In the case that perfect CSI is assumed and 1-bit quantization digital to analog converter(ADC)is used in the BS,EP based detection scheme4 and LMMSE based scheme 5 are proposed for multiuser signal detection with1-bit quantization.By jointly estimating the users signal and the received signal,the proposed detection schemes 4 and 5 achieve significant performance gain over the conventional LMMSE estimator.Secondly,sparse channel estimation is investigated in massive MIMO broadband wireless communication systems.In the downlink sparse channel estimation of massive MIMO orthogonal frequency-division multiplexing(OFDM)systems,the BS-side antennas transmit pilots data on the same time-frequency resources.It is worth noting that,the massive MIMO-OFDM systems work in the frquency disvision duplex(FDD)and the downlink sparse channel estimation is carried out at the users.Then,an approximate message passing(AMP)and sparse Bayes learning(SBL)based channel estimation scheme 1 is proposed by using the spatiotemporal joint sparsity of the system channel.Simulation results show that the channel estimation scheme 1 can approach the optimal performance given by Oracle least squares(LS).In hybrid beamforming(BF)massive MIMO systems operating in time division duplex(TDD)mode,the channel estimation is carried out in the BS.By using the sparsity of the flat fading channel in the angular domain and the priori information of the support,a channel estimation scheme 2based on matching filter(MF)and AMP-SBL algorithms is proposed.Simulation results show that the training data overhead can be greatly reduced based on the channel estimation scheme 2.Then,multiuser signal detection is investigated in the uplink grant-free NOMA systems.For the OFDM-NOMA system whose BS has the perfect CSI,by making full use of the priori information of the user signal and the frame-wise joint sparsity of user activity,a NOMA detection scheme 1 is proposed based on AMP and expectation maximization(EM)algorithms.Simulation results show that the NOMA detection scheme 1 is better than the traditional CS based NOMA detection algorithms,which is due to the fact that the traditional CS-based NOMA detection algorithms can not exploit the prior information of the discrete users signal.For the single carrier(SC)NOMA system,and by using the block sparse distribution of channel vector,we propose a NOMA detection scheme 2 based on EP algorithm,which can jointly estimate the user activity and the corresponding CSI.Simulation results show that NOMA detection scheme 2 has better detection performance than that of the block orthogonal matching pursuit(BOMP)algorithm.Finally,how to improve the IM detection performance of SCIM(Single-Carrier Index-Modulation)systems is studied.This can be achieved by further increasing the path diversity gain of the system.Based on the idea,we proposed a grouped transmission SCIM(G-SCIM)system,which could transmit the same IM information at the successive J symbol periods,but in the form of J different IM signals.Then,based on AMP,we propose a joint-AMP algorithm for the detection of GSCIM systems.Simulation results show that compared with the traditional SCIM systems,the proposed G-SCIM systems show a significant performance gain in terms of IM detection.