Research On Channel Estimation Algorithm In High Mobility Scenario

For the past few years,with the rapid expansion of high-speed rail system and highway network,relevant researchers have continued to increase their attention to wireless communication technologies in high-mobility scenarios.In the high-mobility scenario,complex and changeful environment will make the wireless channel is showing a fast time-varying characteristic and non-stationary characteristics,making timely transmission process cannot obtain accurate channel state information.This poses a very high challenge to signal equalization and coherent detection at the receiving end.This thesis mainly studies the channel estimation problem in a high-mobility environment.The research content is as follows:1.In the high-mobility scenario,this thesis proposes a channel estimation network based on the principle of channel reconstruction and recovery,namely CRD-Net.Aiming at the problems caused by the non-stationary and fast time-varying characteristics of wireless channels,the CRD-Net first uses the channel reconstruction network to model the channel by interpolating to obtain the initial estimate of the channel matrix,and then uses the recovery network to further reduce the influence of channel noise for improving estimation accuracy.The simulation results of single-input single-output orthogonal frequency division multiplexing system show that the CRD-Net based on the principle of channel reconstruction and recovery has lower normalized mean square error in both the frequency domain and the time domain compared with the traditional channel estimation algorithm,and effectively reduces the computational complexity of channel estimation.However,the proposed network does not thoroughly consider the noise abnormalities when the inter-carrier interference caused by the doppler frequency shift is equivalent to the influence of noise,which will bring the poor performance of the normalized mean square error in the case of low signal-to-noise ratio.2.All mention of CRD-Net based on network,put forward the improvement scheme based on enhancement denoising,namely the CRED-Net.The new scheme uses different signal-to-noise ratios to calculate the noise level value of the current channel matrix,and uses the noise level value as input data for learning,which can effectively solve the problems caused by noise abnormalities.In order to reduce the computational complexity,the channel matrix is reduced by downsampling to reduce the training parameters,and the attention mechanism is used to improve the feature expression ability of the network.Through the simulation verification in single-input single-output and multiple-input multiple-output orthogonal frequency division multiplexing system,the proposed improved network based on enhanced denoising shows lower normalization mean square error compared with CRD-Net and traditional channel estimation algorithms,furthermore a performance improvement has been achieved.