Research On Modulation Recognition Technology Of Digital Communication Signal Based On Multidimensional Characteristics

Modulation pattern recognition is an extremely important technology in the field of non-cooperative communication.It focuses on the research of what kind of communication scheme the received radio signals belong to,so as to provide a basis for the subsequent acquisition of communication information and interference with the sender.It plays an important role in the field of military communication countermeasures and the field of civil electromagnetic spectrum regulation.In this paper,eleven kinds of digital signals are studied,including 2ASK,4ASK,8ASK,2PSK,4PSK,8PSK,2FSK,MSK,16QAM,32QAM and 64QAM.In the pre-processing stage of modulation recognition,by analyzing the characteristics of each signal in the time domain,frequency domain and cyclic frequency domain,a scheme for detecting the existence of communication signals based on the spectral peak characteristics at non-zero cyclic frequency of cyclic spectrum is designed.In an environment with the signal-to-noise ratio?SNR?of-4dB,the correct detection probability of all signals can reach 100%.Aiming at different modulation modes,this paper takes the characteristics of instantaneous information,cyclic spectrum and high-order cumulant as the entry point,presents the difference of signals in multi-dimension,and extracts nine kinds of characteristic parameters for modulation identification.In order to solve the problem that MSK signal is difficult to recognize directly,based on the statistical results of instantaneous information,a new feature named nonlinear phase peak factor is proposed in this paper.The experimental results show that MSK signal has 95%correct recognition probability at the SNR of 12 dB.In addition,on the basis of studying the envelope difference of the amplitude variance of the cyclic spectrum cross section of f?28?0,the differential biquadratic processing of the signal is carried out in advance,and the characteristic parameters₂Yis constructed to effectively distinguish the MQAM signals from the MPSK?M=4,M=8?signals in the low SNR environment.In this paper,the decision tree?DT?classifier and the support vector machine?SVM?classifier are designed respectively by using nine kinds of multi-dimensional characteristic parameters.Also,automatic modulation recognition of eleven kinds of digital signals is completed.Aiming at the problem that the correct recognition rate of a single classifier is not high in low SNR environment,this paper deeply studies the AdaBoost ensemble algorithm,and completes the simulation of AdaBoost-DT and AdaBoost-SVM ensemble algorithm under the circumstances of fixed SNR using single-level decision tree and support vector machine as weak classifiers respectively.The simulation results of the two ensemble algorithms are compared with those of traditional decision tree and SVM.Experiments show that the integrated classifier based on AdaBoost algorithm is more accurate than single classifier in recognition performance.It has the advantage of effectively improving the signal recognition probability of low SNR conditions.