Broadband Frequency Hopping Signal Acquisition Technology Based On Time-Frequency Matrix Detection

In the field of wireless communication,frequency hopping communication is a communication method that has developed rapidly in recent years.Different from the traditional fixed-frequency communication method,its biggest feature in communication is that the frequency of the signal will keep jumping.Therefore,it has excellent concealment performance and anti-interference ability,and it is often difficult to effectively capture the frequency hopping signal in actual monitoring.At the same time,with the advancement of wireless communication technology,the electromagnetic environment is complex and changeable,and the signals received by monitoring often contain various interferences.Therefore,it is of great significance to effectively and accurately detect the frequency hopping signal and determine whether the time-frequency overlap is caused by the influence of the interference signal.Based on the time-frequency transformation of the signal,combined with the deep learning target detection theory,this paper studies the problem of frequency hopping signal detection.The specific contents are as follows.(1)This paper summarizes several time-frequency transformation methods commonly used at present,studies their transformation principles,and compares and analyzes them in terms of timeliness and transformation performance.At the same time,this paper studies several classical deep learning target detection algorithms,and sorts out the detection process,which provides a basis for the design of the frequency hopping signal detection algorithm.(2)Aiming at the difficulty of enemy frequency hopping communication signal detection,this paper proposes a wideband frequency hopping signal acquisition technology based on time-frequency matrix detection by combining deep learning target detection and signal time-frequency transformation methods.Firstly,time-frequency transform is performed on the frequency hopping signal,and the transform result is normalized and cut to generate a training data set.Then build and train a deep learning target detection network,use the residual shrinkage unit to improve the robustness of the network,and at the same time reduce the complexity of the network and improve the operation speed of the network;Finally,the trained network is used to detect the frequency-hopping signal,and the occurrence time and frequency range of each hopping signal is obtained.(3)Aiming at the situation that our frequency hopping communication is disturbed and the communication status cannot be fed back,this paper improves the algorithm basis on(2).It can not only detect each hop of the frequency hopping signal but also determine the type of the interfered signal,which realizes the detection of the frequency hopping signal in the case of time-frequency overlap.At the same time,considering that it is difficult to obtain a large number of disturbing data samples in practical applications,data enhancement methods such as Mix Up and Cut Mix are used to expand the training data set to ensure high robust detection performance in the case of small samples.