Research On Intrapulse Analysis Technology Based On Instantaneous Frequency

Intra-pulse analysis is an important part of signal processing in electronic countermeasures.By extracting features from the captured enemy radar pulses,it is possible to recognize the modulation type and estimate their parameters.Then we can make effective reconnaissance and jamming behaviors for specific modulation signals to gain the initiative in electronic warfare.However,with the rapid development of wireless communication technology,the types of modulation signals are increasing,and their parameters have become more complicated,which undoubtedly makes it difficult to recognize modulation methods.As an important development direction of artificial intelligence,deep learning has made breakthrough progress and been widely applied in many fields such as voice recognition,image recognition and image processing.Therefore,in order to improve the performance of signal recognition and parameter estimation with low signal-to-noise ratio,combining deep learning with intra-pulse analysis technology has become an inevitable trend in the development of this field.In this paper,first,a signal separation method is proposed to solve the problem of overlapping signals in the time domain.Then,combined with the instantaneous frequency characteristics of the signal,a modulation recognition method based on deep learning is proposed.The specific content of this paper is summarized as follows:1)Several commonly used parameter estimation algorithms are introduced,including signal-to-noise ratio estimation,sine wave frequency estimation,and symbol width estimation.For modulation signals such as MPSK,LFM,Costas,the phase expansion algorithm and the instantaneous frequency estimation algorithm with low signal-to-noise ratio are introduced,and the instantaneous characteristics of modulation signals are studied in this paper.2)Aiming at the inconsistency of noise addition standards,the advantages and disadvantages of two different noise addition methods are analyzed by simulation in this paper.3)Aiming at the problem of overlapping signals in the time domain,a signal separation method based on the reconstruction filter bank is proposed in this paper.The signals are separated and reconstructed by adopting the steps of channel separation,pulse extraction,channel fusion and signal reconstruction.The above steps are explained and simulated,and the implementation and optimization methods are briefly introduced in this paper.4)For a variety of modulation signals,in order to improve the recognition performance with low signal-to-noise ratio,a signal recognition method based on convolutional neural network(CNN)is proposed,which combines CNN with the instantaneous frequency.Eleven modulation signals such as LFM,MPSK and Coastas are selected.The two-layer CNN network and GooLeNet network are used to recognize the instantaneous frequency images of the signals.By adjusting the parameters such as the learning rate of the model and the size of the convolution kernel,the recognition performance has been significantly improved.Therefore,for the recognition of modulation signals,it provides a certain reference for the optimal model design of convolutional neural networks.