Research On Modulation Pattern Recognition Method Of Communication Signal Under Impulse Noise

Automatic Modulation Classification(AMC)is a commonly used signal processing method,which is used in a variety of different communication fields.It uses presetcriteria to deal with communications with different interference and noise in a given signal set.The signal modulation type of the signal is judged.At present,most methods of communication signal recognition are based on the modulation method recognition based on additive noise that obeys Gaussian distribution.But,recently,many cscholars have discovered that most of natural noise,man-made noise,howerver,Alpha stable distribution model can more accurately represent the above impulse noise.This type of noise does not obey the Gaussian distribution,and the Alpha stable distribution model can more effectively represent the above-mentioned non-distribution.Gaussian noise.Because of the particularity of impulse noise,conventional feature expression methods can not characterize it,so many communication signal modulation recognition algorithms in Gaussian noise environments cannot be directly applied in non-Gaussian environments.Therefore,exploring communication signal recognition methods based on impulse noise environm ent has important research value and significance.The purpose of this paper is to study the modulation pattern recognition algorithm of conventional communication signals under the background of impulse noise,and to in-depth study the modulation principle and recognition method of communication signals.The main work of this paper includes:First,the basic principles of communication signal modulation types and the Alpha stable distribution that can describe impulse noise are researched and analyzed.The definition,parameters and properties of the Alpha stable distribution are introduced in detail,which is the recognition of the communication signal modulation pattern based on impulse noise used in this article.The method provides a theoretical basis;secondly,because the traditional time-frequency transform method degrades or even fails to express the characteristics of the signal after impulsive noise is added,this paper adopts the theory of fractional low-order short-time Fourier transform,and adopts synchronous extrusion and The combination of spectrum rearrangement and the corporate method improves the time-frequency energy divergence problem of the STFT time-frequency method.Simulation experiment analysis shows that this method can well express the time-frequency characteristics of modulated signals under impulse noise.Finally,the optimized lightweight neural network is combined with the obtained time-frequency features,and the obtained time-frequency image of the modulation signal is used as the feature set of the lightweight neural network to train the network.The simulation experiment shows that the algorithm in this paper has good recognition performance in the background of Gaussian and impulse noise in the environment of impulse noise.