Research On Electromagnetic Signal Recognition Technology Based On Machine Learning

With the continuous development of modern communication systems,the wireless signal environment becomes increasingly complex and the demand for wireless spectrum resources is gradually increasing.However,there is a serious imbalance in the allocation of wireless spectrum resources at this stage.Some spectrums are severely overloaded and there are a lot of cross-technology interference problems.Some spectrums are occupied by illegal signals on a large scale,which interferes with the normal transmission of legitimate signals and seriously affects the communication quality.Therefore,the regulation of spectrum resources and the use of radio frequency spectrum are urgent problems to be solved at this stage and even the next-generation communication systems.In a wireless environment which contains multiple electromagnetic signals,the detection and identification of suspicious signals in a certain frequency band is the prerequisite and basis for wireless spectrum supervision.This paper focuses on the subject of electromagnetic signal recognition.It is mainly based on machine learning methods.From the aspects of signal detection and parameter identification,this paper has completed the identification of electromagnetic signal types contained in unknown wireless signals and the identification of Quadrature Amplitude Modulation(QAM)symbol modulation methods adopted by communication signal and the estimation of radar signal modulation parameters.First,based on time-frequency analysis algorithm and target detection technology,this paper proposes a kind of electromagnetic signal type recognition method.For the unknown wireless signal,the method obtains the time-frequency image of the signal through time-frequency analysis in advance.Then it creatively uses You Only Look Once(YOLO)to detect the type of electromagnetic signal existing in the wireless signal and acquire the frequency and time position of the specified signal.Experiments show that the method has good detection effect on mixed communication signals and radar signals,simultaneously has a certain degree of anti-noise performance.After completing the signal type identification of unknown signals,in order to further identify the QAM symbol modulation method of communication signal,this paper has designed a system for QAM method recognition of Orthogonal Frequency Division Multiplexing(OFDM)signal based on machine learning.This system innovatively proposes a ratio-based signal preprocessing algorithm,which can eliminate channel interference caused by signal transmission.Finally,the effectiveness of the pre-processing algorithm and recognition system is verified by comparison experiments and actual experiments.After completing the signal type identification of unknown signals,in order to further identify the radar signal modulation parameters,based on the image processing algorithm,this paper proposes a method that can be used to estimate the radar signal pulse duty cycle and frequency modulation slope parameters.This method can complete the estimation of radar pulse duty cycle and frequency modulation slope by processing the time-frequency image of radar signal.Experiments prove that the method is still feasible in the presence of noise interference.