Research On Detection Algorithm Of Spatial Modulation Receiver Based On Deep Learning

Spatial modulation(SM)is a new novel Multiple-Input Multiple-Output(MIMO)technology,it uses antenna indices to carry bit information,that is,only one transmit antenna is activated in any time slot.The receiver recovers the corresponding antenna index and the modulation symbol,thereby demodulating the bit information.Therefore,spatial modulation technology can achieve a good compromise in terms of reliability,energy efficiency,spectral efficiency and complexity,and has become one of the key technologies of the next generation wireless communication system.The receiver detection algorithm is the decisive factor for the complexity and performance of the whole communication system.Especially in the spatial modulation system,the receiver must estimate the transmit antenna index and recover the modulation symbols.So the design of the detection algorithm is very important.Among the existing detection algorithms,the Maximum Likelihood(ML)detection algorithm based on exhaustive search can guarantee the best performance,but the complexity is also the highest.The maximum ratio combining(MRC)detection algorithm is less complex,but the performance is sub-optimal;in addition,many detection schemes are based on the principle of compression sensing,or based on the sphere decoding design,to obtain a compromise between complexity and performance.Different from the existing research work,this thesis uses the neural network model to design the receiver,and builds the anti-noise deep neural network model to improve the reliability of the spatial modulation receiver detection,while taking into account the system complexity.The main research work of this paper is as follows1.Detection algorithm for space shift keying(SSK)based on deep denoising autoencoder:Space shift keying is the most classical spatial modulation design scheme.This scheme only uses antenna serial number to carry bit information,and the receiver needs to recover the corresponding antenna number.In this paper,a deep anti-noise auto-encoder is constructed to reconstruct the receiver with Gaussian white noise data.In order to balance the time complexity of detection,this paper groups the reconstructed data in the dimension,and uses the mean of the distance within the group to propose an adaptive threshold,and gradually narrows the search space to improve the search efficiency while ensuring the search accuracy.Simulation results show that the model has good denoising performance,the model convergence effect is good and stable,and it has good robustness under different antenna configurations and different SNR conditions2.Detection algorithm design for spatial modulation based on deep denoising neural network:The receiver of the spatial modulation system estimates both the antenna index and the modulation symbol.In this thesis,the complex signal and modulation symbol are first decomposed into real data,and a deep anti-noise neural network is constructed.The noise-bearing data is used as the input of the model,and the index number and modulation symbol of the antenna are combined to obtain the corresponding label.Model training is carried out.This model trains the model end-to-end by optimizing the joint loss form of classification error and reconstruction error,which reduces the impact of noise on network performance.The simulation results show that the detection algorithm based on deep anti-noise neural network is robust under different antenna number settings and different SNR conditions,which can significantly improve the reconstructed data SNR and achieve better performance than MRC.