Research And Design On Blind Detection For Spatial Modulation MIMO Systems

Spatial modulation(SM)with reduced radio frequency(RF)chains,can make a tradeoff among spectral efficiency,energy efficiency,and system complexity,thus being one of the promising multiple-input multiple-output(MIMO)techniques for future wireless communications.Most existing studies on spatial modulation are based on the ideal assumption that the perfect channel state information(CSI)is available.Typical detection algorithms for spatial modulation,such as matched filter(MF)detection and maximum likelihood(ML)detection,are based on the CSI at transmitter.In actual situations,however,it is very difficult to obtain accurate estimation of CSI and the training overhead for that can be costly.For example,channel modeling is pretty difficult in some complex communication scenarios,and efforts should be made to reduce the pilot overhead of energy-constrained nodes in the wireless sensor networks(WSN)or Internet of Things(IOT).Against the background,with the aid of machine learning and link adaptive techniques,this thesis investigates semi-blind detection schemes with low training overhead and the differential blind detection schemes without training overhead,in order to improve the transmission reliability of SM MIMO systems.The main contributions and innovations in this thesis are summarized as follows:(1)The machine learning-aided semi-blind detection schemes for spatial modulation are proposed.Using the training sequence with a limited length,the transmitted signal recovery is achieved without explicitly estimating CSI.The proposed schemes can improve the bit error rate(BER)performance.In the proposed schemes,a transmission frame within the coherence time is divided into two phases,namely the training phase and the data transmission phase.Thus,we can obtain the mapping function between the input and output of the channel through online training.Firstly,the approximate equivalence between the blind detection problem and the clustering problem of the generalized space shift keying(GSSK)system is proved by theoretical derivation,and then we can solve the blind detection problem using the improved k-means clustering(IKMC)algorithm.Secondly,the number of activated antennas increases in the GSSK system causing performance degradation of the IKMC detector.To deal with the problem,we propose a multiclass k nearest neighbor(KNN)based semi-blind detector and a multiclass support vector machine(SVM)based semi-blind detector by considering the semi-blind detection problem as a multiclass classification problem.The simulation results show that the three machine learning-based semi-blind detection algorithms can obtain BER performance gains,while they maintain low complexity.(2)An adaptive semi-blind transmission mechanism based on clustering for SM systems is proposed.By jointly considering the uplink and downlink transmissions of time division duplex(TDD)SM MIMO systems,we design an adaptive semi-blind transmission mechanism based on clustering for SM systems with the aid of adaptive antenna selection.In the proposed transmission mechanism,IKMC detector is adopted to perform semi-blind detection for both the uplink and downlink transmissions.With channel reciprocity and the results of clustering-based detection for uplink transmission,CSI required for adaptive downlink transmission is easily obtained.From the perspective of the overall communication systems,the transmission mechanism improves reliability with very low training overhead.Simulation results show that,compared with the scheme without adaptive design,the proposed scheme can obtain a significant BER performance gains.Besides,the performance of the proposed scheme is close to the optimal adaptive scheme with perfect CSI.(3)Differential beamspace modulation(DBM)blind detection based on beam switch is proposed.For uplink transmission of the TDD MIMO systems,we propose a blind detection scheme based on DBM for transmission over quasi-static channels.The differential beamspace modulation maps information bits onto the beam index and the activation order of beams in the beamspace through beam switching.Firstly,we design the DBM symbols according to the set closure.Then the precoding matrix is constructed based on singular value decomposition(SVD)and the detailed process of DBM transmission is described.Finally,the simplified formula of DBM blind detection is derived.In addition,the theoretical and actual spectral efficiency of DBM are compared.Simulation results show that,compared with differential spatial modulation(DSM),the proposed DBM blind detection can achieve significant performance gains.