The Study Of The Method For Emitter Fingerprint Feature Extraction In Underwater Acoustic Communication

Information warfare plays an more and more important role in the modern military field.As a result,in the field of underwater acoustic warfare,technique for detecting and identification of underwater targets has been a hot spot of studying.Specific emitter identification is an arisen technique for object identification recently,which realizes the classification and identification of enemy's communication devices and weaponry by capturing the emitter signals and extracting the fingerprint characteristic from signals.It is significance to seizing the initiative in the war.This paper studies the method for emitter fingerprint feature extraction in underwater acoustic communication.Considering the non-stationary and nonlinearity of emitter signal,this paper proposes an algorithm for feature extraction based on wavelet entropy and chaos theory.The main research works are as follows:(1)Analyzes the research background and significance of emitter fingerprint feature extraction and individual recognition.Challenge for the studying of emitter fingerprint feature extraction in underwater acoustic communication is also analyzed.investigates the research status domestic and abroad.(2)Considers the nonlinear effect of power amplifier as the main source of emitter fingerprint feature,analyzes the nonlinear distortion and memory effect of power amplifier.What's more according to the application scene of underwater acoustic communication,the influence of power amplifier to OFDM signals is analyzed.Establish the Memory Polynomial model,different individuals can be got through setting different model parameters.And the AM/AM and AM/PM characteristics were analyzed to verify the validity of the model.(3)Proposes an algorithm for emitter fingerprint feature extraction based on wavelet entropy and manifold learning.Firstly,emitter signal is decomposed to several sub-band by wavelet packet transformation,to extract the signal's nonlinear essential features and restrain the influence of in-band noise,the sub-band signals are dimensionally reduced using local tangent space alignment(LTSA)algorithm.The feature vector of wavelet entropy is got by calculating the information entropy of sub-band signals after dimensionality reduction.Simulation result verifies the algorithm's effectiveness for distinguishing different emitter.And simulations are performed to analyze the stability of wavelet entropy characteristic with the variations of the bandwidth and sampling frequency of signal.(4)Proposes a method for emitter fingerprint feature extraction based on chaos theory.Reconstruct the phase space of emitter signal,and then calculate the correlation dimension,the largest Lyapunov exponent and the Kolmogorov entropy based on the analyzing of the phase space.Calculate the Hurst exponent by analyzing the time series of emitter signal using R/S method.Simulation result verifies the four chaos characteristic parameters' similarity among the emitter signals of the same kind and their effectiveness for distinguishing different emitter.And simulations are performed to analyze the stability of chaos characteristic parameters with the variations of the bandwidth and sampling frequency of signal.(5)Perform the classification and identification of different emitter by simulation and experiment.Artificial neural network is used as classifier.Combine the extracted feature vectors of wavelet entropy and chaos characteristic,input the joint feature vectors of different emitter to multilayer perceptron,calculate the rate of identification in different SNRs and situations in which signals are modulated in different ways.Results show that the methods for emitter fingerprint feature extraction proposed in this paper are effective.