Automatic Recognition Of Modulation Signals Based On Instantaneous Information And Spectral Features

Modulation type is an important feature of communication signals.The automatic identification of modulation way means that the modulation type of received communication signals is judged on the premise of no manual intervention.The technology of automatic identification of communication signals is key for software radio and non-cooperative reception,which plays an important role both in military and civil fields.In recent years,with the rapid development of wireless communication technology,the modulation types have become more and more complicated.In addition,the electromagnetic environment of signal transmission has become increasingly poor.These changes make it more and more difficult to identify the modulation mode of communication signals.This paper focuses on the key technologies involved in the automatic identification of modulation types in non-cooperative communication environment,including the detection of the existence of communication signals,the estimation of key parameters,the extraction and selection of characteristic parameters,and the design of classifiers.The set of communication signals to be recognized is {2ASK,4ASK,2FSK,BPSK,QPSK,8PSK,CPM,LSB,USB}.First of all,based on the deep research on the related theory of automatic identification technology of communication signals,this paper focuses on the part of modulation identification preprocessing,starting respectively from the frequency domain and the cycle frequency domain,two presence detection algorithms of communication signals are studies.For the content of parameter estimation,four common carrier frequency estimation algorithms and one blind signal-to-noise ratio estimation algorithm based on subspace decomposition are studied.Simulation results show the effectiveness of the algorithms.Secondly,aiming at the feature extraction part of modulation recognition,by analyzing the instantaneous information,power spectrum and high order spectral features of the signals,a new modulation pattern recognition algorithm based on instantaneous information and spectral features is proposed.The algorithm refers to three classical characteristic parameters,in which the classical characteristic parameter used in the intra-class ASK signals recognition is optimized,and the recognition performance of 2ASK and 4ASK under the condition of lower signal-to-noise ratio and symbol imbalance is improved.Aiming at the shortcomings of 2FSK recognition in the past research,a new identification feature is proposed based on the power spectral density characteristics of different modulated signals,which effectively solves the problem of 2FSK and CPM?PSK signal identification under low signal-to-noise ratio.The algorithm are applicable whether the power spectrum of 2FSK presents a single or double peak.Then,by analyzing square spectrum and quadratic spectrum features of PSK and CPM signals,two higher-order spectral characteristic parameters are constructed respectively for BPSK and QPSK signal recognition.Aiming at the shortcomings that the algorithm of the 8PSK and CPM signal recognition is greatly influenced by the noise in the existing research,the correspondence between the nonlinear phase and the instantaneous frequency of the modulated signal is analyzed.According to whether the instantaneous frequency of the modulated signal appears peak at the symbol switching moment,a new feature parameter for identifying 8PSK and CPM signals is proposed.Compared with the previous algorithm,this algorithm is suitable for CPM signals of any modulation parameters.Finally,in the part of modulation recognition classifier,decision tree and BP neural network classifier are respectively used to judge the signals to be recognized according to the extracted characteristic parameters,and the automatic recognition of nine kinds of communication signal is realized.The correct recognition rate of the algorithm is calculated.Simulation results show that the overall correct recognition rate of the decision tree classifier is 100% when SNR is no less than 9dB,and the overall correct recognition rate using BP neural network classifier is over 95% when SNR is no less than 15 dB.