Font Size: a A A

Research Of Multi-dimensional Fingerprint Feature Extraction And Fusion For Specific Emitter Identification Technology

Posted on:2019-12-04Degree:MasterType:Thesis
Country:ChinaCandidate:Y S XuFull Text:PDF
GTID:2428330611493600Subject:Army commanding learn
Abstract/Summary:PDF Full Text Request
In recent years,the research on specific emitter identification(SEI)has attracted much attention,which is of great significance in the battlefield situation analysis and communication countermeasures.The fingerprint feature which has the measurability,the independence and the uniqueness takes the individual radiation source signal as the carrier.SEI technology based on the characteristic which the fingerprint has,achieves the goal to classify the radiation source.Firstly,this paper starting from the individual fingerprint of the radiation source,analyze the mechanism of the emitter signal distortion and establish mathematical models.Based on the models,the emitter signal is simulated.Secondly,a fingerprint extraction method based on constellation error is proposed on the basis of emitter signal distortion.Refer to the fingerprint extractions based on HHT spectrum and bispectrum respectively,a extraction based on the improved intrinsic time-scale decomposition(IITD)method is proposed.The above fingerprint extraction methods are verified by using the simulated signal data and the measured ship signal data.Thirdly,the dimensionality reduction methods and feature selection methods are used to screen high-order fingerprint feature data.The data dimension reduction methods include principal component analysis(PCA)and Fisher linear discriminator(LDA).The feature selection methods adopt MIFS,mRMR and JMIM are studied.These data processing methods can solve the problem of high dimensional and improve the individual recognition performance.At last,the identification of multi-receiver emitter individuals based on D-S theory is studied.Information fusion based on feature level avoids serious data loss on one hand,and on the other hand has a nice fusion effect.According to the rectangle of the fingerprint features,the basic probability assignment of the fingerprint features is proposed by using interval numbers.The resolution method of D-S evidence conflict is explored to solve the problem of evidence credibility between receivers.
Keywords/Search Tags:SEI, Fingerprint mechanism, Fingerprint features extraction, Feature selection, Multi-receiver information fusion based on feature level
PDF Full Text Request
Related items