Research On Modulation Identification Method Of Digital Signals Based On Time-frequency Analysis

Modulation pattern recognition of signals is a key technology of non-cooperative communication scenarios and cognitive radio systems.Therefore,it is a key problem to be solved in this field to find an algorithm that can effectively mine the essential characteristics of signals,realize automatic recognition of modulation mode with different recognition algorithms and classifiers,and achieve the required recognition accuracy in various communication scenarios.This paper mainly focuses on the current situation that communication equipment such as emergency and fixed stations are subject to more and more uncertain interference.In order to determine the modulation type of interference signals,the digital signal modulation mode identification method based on time-frequency analysis is deeply studied.Firstly,a digital signal modulation mode recognition algorithm based on deep autoencoder is proposed for six kinds of non-stationary digital communication signals.The time-frequency images of the signals are obtained by converting six kinds of digital signals from the time-domain representation to the time-frequency domain representation.Then deep features of time-frequency images are extracted by deep autoencoder,and then multi-layer perceptron classifier is used to identify the modulation mode of the signals.Simulation results show that this algorithm has good robustness and can still effectively recognize six types of digital signals when the signal-to-noise ratio is-10dB.Secondly,to solve the problem that time-frequency distribution images is not enough to distinguish three kinds of MQAM signals,multi-scale wavelet entropy is introduced as the feature of the signals to be recognized and input into BP and ELM neural network classifier to identify the modulation mode of nine kinds of digital signals.In order to further improve the correct recognition rate,the feature enhancement method is introduced to mine the deep features hidden in the original features by considering the high order and interaction terms of the original features.Simulation results show that the feature enhancement method can effectively optimize the original features and improve the classification performance.In the study of the above method,it was found that the features of BPSK and QPSK signals overlapped seriously,and the recognition result was obviously worse than that of other kinds of signals.To solve this problem,a digital phase-modulated signal recognition algorithm based on improved multi-scale wavelet entropy was proposed,which specifically identified the modulation mode of PSK signals.Then a digital modulation signal recognition algorithm based on feature fusion is proposed.The algorithm is divided into two steps.The first step is to distinguish the non-PSK and PSK signals in the nine modulated signals by using the multi-scale wavelet entropy feature.In the second step,7 kinds of non-PSK signals and 2 kinds of PSK signals were further identified.The simulation results show that the algorithm can recognize each kind of signal with high recognition rate,and the overall recognition rate of the signals is improved obviously.