Research On Scenario Recognition Based On Deep Learning In Wireless Communication

With the rapid development of communication technology and the increasing demand for communications,technical solutions that rely on expert knowledge bases and artificial decision rules have gradually exposed the drawbacks of low scalability,and it has become a trend to introduce intelligent methods into communication systems.In the field of wireless communications,the electromagnetic environment is complex and varied,and the signals are diverse and rapidly changing.It is necessary to use intelligent methods to comprehensively identify complex communication environments.In recent years,Deep Learning(DL)technology represented by Deep Neural Networks(DNN)has made great achievements in the fields of image recognition and speech signal recognition.This paper applies deep learning to the problem of wireless communication scene recognition,realizes the intelligent perception of the communication system to different communication environments,and then provides an important reference basis for self-adaptive optimization of subsequent communication schemes.The thesis first proposes a complex model of interference scenario recognition based on Convolutional Neural Network(CNN).The research work in this part includes: performing the time-frequency transformation on the premise that the mixed interfering signal is nonstationary,and then using image processing technology to preprocess the time-frequency image,using the convolutional neural network in deep learning the frequency grayscale image is trained to create an interference scene recognition model,thereby transforming the interference scenario recognition problem into an image recognition problem.Simulation results show that the interference scenario recognition model established in this thesis can achieve high recognition accuracy.The thesis proposes a model of channel scenario recognition based on Deep Belief Network(DBN).The overall performance of a wireless communication system is greatly affected by the wireless channel,so identifying different channel scenarios can play an important role in the design of the communication system.The main work completed in this part is to measure multiple sets of channel data under three different scenarios,use a deep learning correlation algorithm to perform dimensionality reduction analysis,and use a multipath channel estimation algorithm to solve the Channel Impulse Response(CIR).Based on the CIR and received signals,the time domain,frequency domain,and image domain characteristics of the channel propagation environment are extracted.Finally,a deep belief network is introduced to establish a channel scene recognition model.The cross-validation results show that the model realizes the correct recognition of the channel scenario to be predicted.In summary,this thesis applies the convolutional neural network and deep belief network in deep learning to the two issues of hybrid interference scene identification and channel scene identification respectively,and the corresponding models established have high recognition accuracy.The research content of this paper has guiding significance to the future network scene intelligence perception and recognition.