WiFi Interference Source Identification Based On Deep Learning

With the rapid development of wireless short-range communication technology,WiFi communication technology has become more and more mature,and WiFi communication systems are widely used in various fields.However,because WiFi communication system works in the ISM band of 2.4G,which is a public,non-authorized band,it leads to the existence of many other wireless signals in this band,which cause interference to WiFi communication and affect WiFi communication quality.Therefore,to guarantee smooth WiFi communication,this paper researches and analyzes the identification technology of WiFi interference sources as follows:First,this paper analyzes WiFi communication signals,and summarizes and analyzes six common WiFi interference source signals on the 2.4G band,and studies the time-frequency characteristics of WiFi interference source signals through waveform diagrams and spectrograms of the interference source signals to prepare for the subsequent identification work.Secondly,the limitations of the existing conventional signal recognition techniques are analyzed,and combined with the characteristics of the interference source signals to be recognized,it is demonstrated that the existing recognition techniques are difficult to be practically applied to WiFi interference source recognition,for which a WiFi interference source recognition algorithm based on convolutional neural network is proposed,which combines the characteristics of deep learning,the actual acquired signal is fed into the convolutional neural network model for training and learning after pre-processing operations such as modulus taking and normalization,and finally the signal of the interference source to be identified is pre-processed and fed directly into the trained network model,which directly outputs the category information of the interference source.After a large amount of data simulation analysis,it is confirmed that the identification idea based on deep learning is scientifically feasible.Finally,for the shortcomings of the WiFi interference source identification algorithm based on convolutional neural network,two improved identification algorithms are proposed,which are the WiFi interference source identification algorithms based on deep convolutional neural network and residual neural network respectively.The improved algorithm based on deep convolutional neural network changes the data preprocessing operation,replacing the original operation of taking modal values for time-domain data with the operation of separating real and imag data in the time-domain,to maintain the information integrity and retain more original information.Besides,for the problem of insufficient layers of the original neural network model,the number of layers of the network model in the improved algorithm is increased to enable the neural network to learn higher-level and more essential features.The improved algorithm based on residual neural network also implements the same data preprocessing operation,and at the same time,for the problem that the gradient disappears during neural network training and only the higher-level features are used,the idea of residual is used to build a residual neural network model.Finally,the two improved algorithms are simulated and analyzed,and it is verified that the two improved algorithms have excellent and scientifically effective recognition performance under the conditions of only containing interference sources,different JNR and different JSR,and the recognition rate of most interference sources is above 95%,and the recognition rate of some interference sources is as high as 99.9%,and the recognition algorithm based on residual neural network is better than the recognition algorithm based on deep convolutional neural network.