Recognition Of Digital Modulation Signals Based On High-order Statistical Characteristics

The high-order statistical characteristics of digital modulation signals have simple basic principles,high operability,and good suppression of white Gaussian noise.These unique characteristics make higher value in the digital signals modulation recognition.Therefore,whether in theory or in practice,the high-order statistical characteristics of modulated signals are very important.In this thesis,we selecct two statistical characteristics of many characteristics and make a deep discussion.These two recognition methods respectively improve the recognition effect or increase the number of identification signals,which provide a theoretical basis for future analysis and indicate the research direction.The digital modulation signals studied in this thesis are MASK,MFSK,MPSK,MQAM.Based on the theoretical basis,we selecte the representative signals in each modulation type,and analyze the characteristics of signal's time domain,frequency domain and power spectrum.Then this thesis introduces the mathematical significance of higher-order cumulants and higher-order moments,and calculates their correlation formulas.Calculating and constructing a classification decision tree based on the high-order cumulants of the twelve digital modulation signals.The first method uses the characteristic parameters based on the third-order polynomial of signal-to-noise ratio,and optimizes the parameter threshold based on signal-to-noise ratio with a large number of experimental datas;the identification of MQAM signals uses segmented feature parameters and feature parameters whitch are sensitive to signal-to-noise ratio reduction but can solve signals aliasing problems;the signal groups consist of the original signals,the signals after adding the average noise,and the signals after subtracting the average noise,which reduce the erroneous judgment in signal recognition.Next,this thesis introduces the theoretical knowledge of the cyclic spectrum functions and the circular spectrum formulas of the desired signals,uses signal simulations to analyze the characteristics of the cyclic spectrum and calculates the theoretical values of the higher-order cyclic cumulants.The categorization decision tree consists of characteristic parameters based on the high-order cyclic cumulants and the cyclic spectral envelope feature.The second method adds cyclic spectral envelope features(such as variance,mean,etc.),the original method can only recognize MPSK signals,but the new method can increase the recognition of MFSK signals.This method provides theoretical basis for the application of engineering and research direction.Finally,the experimental results of the first identification method in this thesis show that: the method has better recognition effect than existing single feature parameters,although the signal-to-noise ratio is less than 0 dB,there is still a small amount of aliasing,when the signal-to-noise ratio is bigger than 0 dB,the signals can be effectively identified.The experimental results of the second identification method in this thesis show that when the signal-to-noise ratio is above 8 dB,the accuracy rate of the algorithm is above 90%.