Research On Modulation Pattern Recognition Algorithm Of Underwater Acoustic Communication Signal

The continuous improvement of the underwater combat network is inseparable from the research and development of underwater communication network security and underwater information countermeasure technology.Modulation pattern recognition for intercepted non-cooperative communication signals is the key stage of investigation in the confrontation phase.After obtaining the modulation mode of the intercepted signal,the signal can be estimated and demodulated.Also the signal characteristics of the enemies can be simulated to perform fraudulent interference.The modulation pattern recognition technology of radio communication signals has been relatively mature,but due to the influence of underwater channel time-varying,space-variant,high-frequency attenuation and serious Doppler,the direct application to underwater acoustic communication signals is not effective.How to find a high-performance pattern recognition algorithm adapted to different underwater communication environments has become a new focus in the field of underwater communication identification.Firstly,this thesis introduces the basic principles and basic processes of pattern recognition and studies the methods of signal preprocessing and feature extraction for underwater acoustic communication Experimental analysis is carried out for several feature extraction methods,and the instantaneous characteristics of the signal,the statistical parameters based on the statistics.the high-order accumulation and the time-frequency characteristics based on the short-time Fourier transform and wavelet transform are extracted.Secondly,due to the instability and randomness of the harsh underwater acoustic channel,the modulation pattern recognition algorithm has higher requirements on the performance of the classifier.The research from the classical K-nearest neighbor classification to the deep learning classifier algorithm is carried out.Finally,the feature extraction and classification recognition algorithms are combined to study the K-nearest neighbor and CART decision tree recognition algorithm based on feature parameters and high-order cumulative series features,and the short-time network classification algorithm based on instantaneous features.Time-frequency based TLGoogLeNet convolutional neural network underwater acoustic communication signal pattern recognition algorithm completes three major types of 5 kinds of modulation methods pattern recognition.They are single-carrier modulation(BPSK,QPSK8PSK),direct sequence spread spectrum(DSSS)and orthogonal frequency division multiplexing(OFDM).Through simulation experiments,the feasibility of each recognition algorithm is verified and the influence of signal-to-noise ratio on the accuracy of various recognition algorithms is analyzed.Through the verification of underwater communication test data,the performance of the proposed algorithm is evaluated and analyzed.Among them,the accuracy,recall,and the F1 values of the TLGoogLeNet-based convolutional neural network pattern recognition algorithm are all more than 97%,which successfully realizes the accurate identification of the five modulation signals.This thesis can provide the necessary technical support for the underwater acoustic communication countermeasure system.