Modulation Recognition And Parameter Estimation In Complex Electromagnetic Environment

Modulation recognition is an important research area in the field of communication.This technology has been widely used in civil and military communication such as signal detection,interference signal recognition,and electronic countermeasure.In the civilian field,with the development of communication technology,modulation methods have become more diverse,spectrum resources have been increasingly crowded and overlapping,and background noise and interference have increased remarkably;in the military field,especially in electronic countermeasure,signals propagation may be directly affected by the energy of the electromagnetic signal,and so is the communication channel.Therefore,in order to ensure the smooth reception and content recognition of the signal,it is necessary to study the modulation recognition of the signal in a complex electromagnetic environment.Modulation recognition problem is a typical pattern recognition problem.The recognition algorithm is mainly based on two methods: maximum likelihood decision and feature extraction.In this thesis,the modulation recognition method of signals and the carrier frequency estimation of multiple phase shift keying algorithms are studied.The main work and contributions are as follows:First,when the modulation type is identified using the transient characteristics,the recognition rate is low at low SNR,and the complexity is higher when using high-order cumulant combination parameters.For the problem,a dynamic decision method based on SNR is proposed.The dynamic identification method takes advantage of the transient characteristics when the SNR is higher than the threshold;otherwise,the combined values of the high-order cumulant are calculated in order to improve the recognition accuracy under low SNR environment.The simulation results show that the recognition accuracy of the algorithm is higher than the method which only uses the transient feature.Because the transient characteristics are directly used when the signal-to-noise ratio is high,the average recognition time is faster than the method that only uses the higher-order cumulant values.Second,for the problem that the characteristic parameters depend on the specific types of the received signal,which needs manual extraction,this paper introduced a neural network-based classifier to complete the modulation recognition of the signal.The neural network-based classifier can perform the extraction of signal features using the network itself other than artificially extraction.The convolutional neural network model and the recurrent neural network model are applied to identify the modulation types of received signals.This thesis replaces the long short-term memory(LSTM)unit with the gated recurrent unit(GRU),which can effectively utilize signal timing sequence information and GRU is simpler than long short-term memory unit structures.The simulation results show that the recognition rate of the GRU model is similar to the rate of long short-term memory model;the time of each epoch of training is significantly shortened.Third,when estimating the carrier frequency of the multiple phase shift keying modulated signal by using the high-order transform method,the accuracy is too low under low SNR.For the problem the paper denoised the received signal with wavelet transform.Since the high-order transform is a nonlinear transform,the noise is further amplified after the transform,which will lead to worse estimating performance.By using the band-pass filter,the spectral component of the signal except for the main lobe is lost,which reduces the accuracy of the frequency estimation.The wavelet transform uses a finite-length attenuated basis function to decompose the signal while removing noise and retaining part of the frequency band information of the signal.The simulation results show that the wavelet transform denoising method has higher carrier frequency estimation accuracy than the traditional band-pass filtering method under low SNR conditions.