Research On MIMO-OFDM Adaptive Transmission Technology Based On Deep Learning

Due to the rapid increase of communication services,how to improve the information transmission rate and communication quality has always been a research hotspot.With the development of artificial intelligence technology,intelligent communication has gradually become the main development trend in the field of mobile communication research.At present,the research in this field is rapidly advancing to the physical layer,in which data-driven network and model-driven network are two representative models in the current intelligent communication research.Adaptive transmission technology can improve the efficiency or reliability of signal transmission by changing the transmission strategy according to various parameters of the received signal.In recent years,the development of deep learning technology has created new opportunities for research in various fields.This paper combines deep learning with adaptive transmission technology to improve the performance of adaptive transmission system,which provides a theoretical basis for the development of MIMO-OFDM adaptive transmission technology.The main work arrangement of the paper is as follows:1.Research on adaptive ensemble deep learning models that can be applied to wireless communication systems.Firstly,the existing typical deep learning models,such as FC-DNN,LSTM,and CNN,are given,and the implementation process of these deep learning algorithms is deduced.Simultaneously,the received power under different path delay is used as the adaptive factor to integrate each sub-network's output adaptively,and the implementation process of the adaptive integrated deep learning algorithm is also deduced.Theoretical analysis and simulation results show that the proposed model can adjust the size of the adaptive factor or transmission strategy according to channel information,and has higher classification performance and adaptability than existing deep learning models.2.Research on the adaptive signal detection scheme of OFDM system based on data-driven network.Firstly,the traditional detection and estimation algorithms,such as LS and MMSE algorithms,are introduced,and the implementation complexity level and BER performance are analyzed.Then,an adaptive signal detection and estimation scheme based on deep learning is given.Besides,the method of feature extraction and label allocation for deep learning model training and the relationship between received power and delay under different paths are described.The signal detection model obtained in the offline training stage is used to complete the online test of the channel estimation and equalization process.Simulation results verify the effectiveness of the scheme,and comparative experiments show that the scheme has better BER performance.3.Research on the adaptive transmission scheme of MIMO-OFDM system based on model-driven network.Firstly,the implementation technology and common criteria of adaptive transmission scheme are introduced,and then an adaptive transmission scheme based on model driven deep learning algorithm is given.The proposed AMNet and ADNet are used to replace the modulation and modulation recognition process of the system.In this paper,channel rank information,channel label information,estimated SNR and equalization information are adopted as feature vectors of AM model,and labels are assigned according to the comprehensive analysis of signal detection performance.Finally,the cyclic spectrum features of the modulation mode are analyzed,and the dimension reduction features are extracted by the spatial transformation ability of CNN to complete the AMR of the received signal.Simulation results show that the adaptive transmission scheme based on model-driven deep learning algorithm has higher data throughput and modulation recognition accuracy in the case of specific BER.