Signal Detection Algorithm Of Wireless Communication System Based On Deep Neural Network

As the rapidly developing of the wireless communication and the Internet in this era,the number of mobile devices which are accessing to the Internet is increasing every day,while people are adapted to the life with mobile Internet,they also put forward higher requirements on the transmission quality and transmission rate of wireless communication system.Multiple Input Multiple Output(MIMO)technology has become one of the key technology of wireless communication system in this new era due to its ability which can greatly improve the spectrum efficiency and transmission rate with no need for additional bandwidth.While the using of multiple transmitting and receiving antennas makes the signal processing more complex at the receiving end,and the complexity will highly increase with the layout of massive antennas.Therefore,studying the signal detection algorithm with low complexity and bit error rate(BER)is still an important research direction of wireless communication technology.Based on deep neural network,the thesis carried out in-depth discussion and research on MIMO signal detection technology.First,the research background and current developing situation of the MIMO technology is introduced,then the thesis analyzes and introduces the characteristic of MIMO channel and system model,and several typical signal detection algorithms of MIMO system and the complexity of the algorithms are briefly analyzed,after that,several typical deep neural network models and their principles are introduced.After analyzing the received signals' category-characteristics of MIMO system,the thesis expounds the connection between category-characteristics of the received and transmitted signals,and the recognition degree of characteristic information under different signal-to-noise ratio is analyzed.Then a signal detection algorithm based on auto encoder(AE)which is used to extract the feature of signal and extreme learning machine(ELM)for feature recognition and classification is proposed,and the simulation results verify the feasibility and good detection performance of the algorithm.To improve the time-effectiveness of the detection algorithm,a detection algorithm based on ELM auto encoder is proposed in the further study,and the selection basis of parameters in the model is analyzed through the experiment.Finally,simulations of serval signal detection algorithms for comparison are conducted,and the simulation results show that the detection algorithm has better BER performance and higher efficiency.