Study On Recognition Method Of Digital Modulation Signal With Non-Gaussian Noise

Modulation recognition of communication signals is the process of determining the modulation of signal with noise from the given signal set and it has important application in military communications and civil communication etc. Noise is assumed to obey Gauss distribution in modulation recognition algorithms of many current literatures. However, artificial electromagnetic noise, natural noise and the combination of both of them are non-Gauss noise in most radio channel. It found that Alpha stable distribution model can effectively represent the non-Gauss noise. In recent years, non-Gauss noise is modeled as Alpha stable distribution model by many scholars. Since the second-order and above second-order statistic of signal is infinity under Alpha stable distribution noise, many modulation recognition algorithm under Gauss noise can not be directly used in Alpha stable distribution noise. So, the study of modulation recognition under Alpha stable distribution noise has important actual value and the theory significance.This paper focuses on the research of recognition of digital modulation signal under Alpha stable distribution noise environment. The main work are:Firstly, aiming at the problem of high computation complexity and poor performance ofthe existing recognition methods of digital modulation signals, a concept of generalizedcumulant and generalized instantaneous phase is proposed and the signals arerecognized based on this concept with Alpha noise. Characteristic parameters which aregeneralized cumulant and maximum of generalized instantaneous phase spectral densitybased on fractional Fourier transform are extracted by this method. And then theminimum mean square error classifier and the threshold are used to achieve modulationrecognition of digital modulation signal with Alpha stable distribution noise. Simulationresults show that the proposed method not only can achieve recognition in low SNR butalso has lower computation complexity than the traditional recognition methods in anAlpha noise environment.Then, a novel recognition method based on transient characteristics and the decision tree classifier is proposed in this paper under Alpha stable noise. The fractional lower order Fast ICA is used to separate the signal and noise and the separated signal is decomposed by on average Local Mean Decomposition which is base Interpolation. Since the instantaneous frequency and amplitude of different types of modulation signal are differnent, two instantaneous characteristics are constructed which are piecewise instantaneous frequency standard deviation and piecewise instantaneous amplitude standard deviation. At last the signal modulation modes are classified by the threshold of decision tree classifier. Simulation results show that the proposed method has the very good recognition performance and robustness in an Alpha stable distribution noise environment.