Research On Detection Algorithm Of UAV Signal Based On Time-frequency Analysis

Frequency hopping signals are usually used in the military field due to their strong antiinterference ability,high confidentiality and low interception rate.As the development of the technology,frequency hopping signal has been gradually applied in the civil field,in which the UAV signal is a typical frequency hopping signal.At present,with the widespread use of the UAV,the phenomenon of black flying is gradually increasing,which will seriously affect social security.Therefore,there has an urgent problem need to solve,that is when the UAV signal under the interference of environmental noise,specialized persons how to quickly capture and identify the UAV signal.This paper’s main research content is to use the STFT & SPWVD timefrequency analysis method to analyze the signal firstly,then use the signal noise reduction based on image processing,generate small sample data set,use the convolutional neural network to quickly classify the signal.Finally,use the Extended Modified B-distribution(EMBD)algorithm to analyze the frequency hopping signal model time-frequency and then combine with the instantaneous frequency method to estimate the signal parameters.Firstly,the frequency hopping signal’s basic principle is introduced in this paper,and the common parameters of frequency hopping signal are introduced briefly.The performance,advantages and disadvantages of common time-frequency analysis methods are analyzed.The Convolutional Neural Networks(CNN)are also analyzed.At the same time,analyzes the principle of fixed hop identification of signal source using power pair elimination method,and the commonly used detection and then introduces the estimation methods of frequency hopping signal.Secondly,this paper proposed the method of STFT & SPWVD frequency analysis,can suppress the noise at the same time,well suppress the cross term,and then use the method based on image processing for signal noise reduction processing,the signal wiener filter processing gray threshold method to eliminate part of the noise,using the morphological filtering method to remove the remaining strong noise.In the low signal-to-noise environment,the signal and the signal features can be enhanced.Then using the image processing method to generate small sample data set,through the network CNN1,CNN2 and BP algorithm for the drone signal classification,before and after classification effect comparison,can further prove the effectiveness of noise reduction method based on image processing,and through the comparison of the three methods of different SNR environment can be concluded that in small sample,the CNN2 method can meet the needs of rapid and accurate identification of UAV signals.Finally,this paper focuses on a new type of quadratic time-frequency distribution-timefrequency analysis method(EMBD)based on extended B distribution correction algorithm,compared with STFT & SPWVD time-frequency analysis algorithm can improve the performance of frequency hopping signal parameter estimation and it also can reduce the amount of computation.Then use the singh window function instead of hyperbolic window.Compared with the EMBD algorithm,the kernel-optimized EMBD algorithm can significantly improve the parameter estimation performance of the frequency hopping signal combined with the instantaneous frequency method.