Message Passing Algorithms Based Channel Estimation And Signal Detection In OFDM With Index Modulation System

Orthogonal frequency division multiplexing index modulation(OFDM-IM)is a novel multi-carrier modulation technique,it only activates parts of subcarriers for transmission and leverages the index of activated subcarriers to transmit the information,and this accounts for the sparsity of OFDM-IM transmission signal.The specific structure that is different from OFDM enables OFDM-IM to have higher energy efficiency,and stronger robustness to Doppler spread these advantages.Firstly,this thesis studies the joint channel estimation and signal detection in time-invariant scenario.On the one hand,the special structure of OFDM-IM that is different from OFDM,i.e.the activation pattern(AP)in each block,increases the difficulty in joint estimation and detection.On the other hand,different message passing algorithms have different application scenarios and virtues and defects,thus how to effectively employ the related algorithms in joint channel estimation and signal detection is also a challenge.Based on the frequency domain channel and time domain channel related signal models,this thesis constructs two corresponding factor graphs and proposes effective message passing algorithms structures,including mean field(MF)-belief propagation(BP)algorithm and pattern nodes related MF-BP algorithm that are based on the frequency domain factor graph,and the BP-MF-BP algorithm and joint generalized approximate message passing(GAMP)-MF-BP algorithm that are based on the time domain factor graph.Among them,the pattern nodes related MF-BP algorithm is the advanced MF-BP algorithm,it considers AP constraint in each OFDM-IM subblock,and thus it shows the best bit error rate(BER)performance with highest complexity.Time domain based algorithms consider the channel time domain-frequency domain transformation,BP-MF-BP adopts BP to address the loopy channel transformation part,thus shows the worst performance.GAMP-MF-BP adopts GAMP to replace BP to address channel transformation part,thus its complexity is the lowest and shows the similar performance as the pattern nodes related algorithms.Secondly,this thesis studies the joint channel estimation and signal detection in time-varying scenario.It considers the signal detection given known channel at first.Due to the fact that time-varying channel induces serious inter-subcarrier interference,the difficulty of signal detection increases.Combining the AP in OFDM-IM and the sparsity of transmitted signal,approximate message passing(AMP)-based detector is proposed.Based on the distribution features of elements in frequency domain channel matrix,a damped AMP detector is designed.Moreover,the number of estimated channel parameters is numerous in unknown channels,thus it is too tricky to jointly channel estimate and signal detect.To reduce the complexity of channel estimator,this thesis adopts basis expansion model(BEM)to estimate channel parameters,a turbo iterative structure is proposed as well.In this iterative receiver,the soft information is iteratively transmitted among the channel estimator,signal detector and decoder: it adopts AMP to estimate the BEM parameters and then reconstructs the channel;it also employs AMP to recover the transmitted signal utilizing the reconstructed channel and soft information from decoder,which is further fed back to the decoder.In addition,damped AMP and vector AMP these two approaches are proposed in channel estimator and signal detector.