Research On Digital Signal Modulation Identification Method Under Alpha Stable Noise Distribution

Traditional recognition algorithm for digital communication signals is usually based on the background of Gauss white noise.While in the real life environment,there are a lot of non-Gauss noise with spike pulse characteristics.In such a channel environment,the traditional modulation recognition algorithm will be invalid.In terms of these noises with spike pulse characteristics,a non-Gaussian noise model based on stable distribution is often used to describe and fit them.And main research contents are shown as follows:Firstly,the theory of the ? stable distribution and non-Gaussian noise in real channel are studied and the definition and properties of stable distribution are given.The non-Gaussian noise in the real channel of the city center is collected,which proves that the non-Gaussian noise in the downtown channel environment in daily life conforms to the stable distribution,and proves that the traditional modulation recognition algorithm will lose its function in the non-Gaussian noise background.Secondly,according to the research status of modulation recognition of digital signal in the ? stable distributed noise background,the fractional low-order cyclic spectrum algorithm is selected as the basis for modulation recognition of digital signal.The basic theory of fractional low-order cyclic spectrum is introduced.The definition of fractional low-order cyclic spectrum and the estimation method of fractional low-order cyclic spectrum are given.Then,taking 2ASK,BPSK and 2FSK as examples,the fractional low-order cyclic spectrum characteristics of several modulated signals are simulated and analyzed.It is proved that different modulated signals have different fractional low-order cyclic spectrum characteristics.Finally,the modulation recognition algorithm based on fractional low-order cyclic spectrum is simulated and optimized.The maximum cross-section projection of fractional low-order cyclic spectrum of digital signal is selected as signal feature,and k-nearest neighbor algorithm is selected as classifier to classify and recognize digital modulation signal.The recognition effect of fractional low-order cyclic spectrum based recognition algorithm and traditional recognition algorithm is compared and analyzed.The relationship between the characteristic index ? of the ? stable distributed noise and the order of fractional low-order cyclic spectrum is further studied.Based on the research results,the overall modulation signal recognition algorithm is optimized,and the recognition rate of the algorithm is improved.