Research On Channel Estimation Algorithm In OFDM System Based On Super Resolution Network

Orthogonal frequency division multiplexing(OFDM)has aroused great interests in wireless communication because of its high-speed data communication ability and robustness to multi-path delay expansion.At present,OFDM technology has been widely used in wireless networks such as long-term evolution technology(LTE)and fifth generation mobile communication(5G),and will be applied to the next generation wireless communication system sixth generation(6G)as a key technology.In the wireless communication system based on OFDM,whether using single antenna or multiple antennas,it is necessary to accurately estimate the channel state information to improve the performance of the wireless communication system,which makes channel estimation very important.According to the research status of channel estimation in OFDM systems,least square(LS)estimation is simple to implement,but its performance is limited.Minimum mean square error estimation(MMSE)significantly reduces the mean square error,but the computational complexity is very high.In order to achieve a balance between estimating performance and complexity,in recent years,with the successful application of deep learning(DL)in various fields,it has attracted wide attention in the communication fields,and promoted people’s interest in using DL to solve communication and signal processing,this dissertation proposes two channel estimation techniques based on deep learning super-resolution networks for single-input single-output(SISO)and multi-input multi-output(MIMO)OFDM systems.(1)In the first research point,the pilot-based channel estimation in SISO-OFDM system is considered.Firstly,the channel response at this location is obtained by transmitting the pilot.Due to the influence of Doppler frequency shift,there is a certain correlation between subcarriers.Using the correlation between channel coefficients at different subcarriers,this paper proposes a channel estimation algorithm based on superresolution generative adversarial network(SRGAN).By building countermeasure network to optimize two-dimensional interpolation,it is easier to learn the correlation and distribution of channel coefficients at different subcarriers in the training process.The simulation results show that the performance of the proposed algorithm is much better than that of LS estimation,and is about the gain of 2.5dB at low signal-to-noise ratio(SNR),and about 5dB at high SNR.At the same time,it is better than linear minimum mean square error(LMMSE)estimation when SNR is lower than 15dB.(2)The second research point considers pilot-based channel estimation in MIMOOFDM systems.Compared with the SISO-OFDM system,the increase of the number of antennas increases the dimension of the channel matrix,which greatly increases the complexity of the channel estimation algorithm.In order to reduce the complexity and ensure the performance of system estimation,a channel estimation algorithm based on self-attention mechanism is proposed in research point 2.By building the self-attention mechanism module,the interdependence between channel coefficients is constructed,so that the complete channel information can be estimated more effectively.The simulation results show that the complexity of the proposed algorithm in second research point is lower than that in first research point,and the estimation performance is similar.The simulation results show that the estimation performance of the proposed algorithm is much better than that of the LS estimation algorithm.compared with the common channel estimation algorithms based on super resolution,it has about the gain of 1.1dB when the SNR is 20dB and the gain of 0.8dB when the SNR is 0dB.At the same time,the performance of the LMMSE estimation algorithm is better than that of the LMMSE estimation algorithm at low SNR and similar to that of the LMMSE estimation algorithm at high SNR.