Research On Feature Extraction Technology Of Communication Transimitter Individual

Individual transmitter identification is a new research direction in the field of communication signal processing.And individual transmitter identification technology in field of the communication against reconnaissance system and wireless network security has a very important application.The main work of this thesis includes the following four aspects from the features of the individual transmitter identification in the field of communication:1.The analysis and extraction of steady-state features based on bispectrums are introduced.Four kinds of integral bispectrums and selective bispectrums were analyzed,and five bispectrums were simulated with the measured signals.The simulation results show that the four integral bispectrums are better than the selective bispectrum.And the recognition rate of the four integral bispectries can reach 80%.2.Realize the detection of transient signals and feature extraction.The starting point and the end point detection of the transient signal are studied.An improved phase detection method is given for the detection of the starting point of the transient signal.On the basis of the phase detection method,the posterior probability density function is used to detect the transient signal starting point adaptively and the measured data of the walkie-talkie are experimented.It is shown that the improved phase detection method have a higher accuracy and sensitivity compared with the other three detection methods.3.The analysis of time-frequency distribution characteristics and polynomial fitting characteristics based on transient signal and three feature extraction methods are studied.The recognition rate of time-frequency distribution and polynomial fitting feature in the experiment is 94% and 78% respectively.Evaluate the separability of features,and the time-frequency distribution feature has better separability in terms of individual identification,but PCA dimensionality reduction is needed before classification due to the time-frequency distribution feature dimension is higher and the PCA needs a lot of calculations.The polynomial fitting feature dimension is small and can be classified directly.On the basis of the separability the index D is obtained by adding the feature dimension information.The evaluation index D of the polynomial fitting feature is the highest and can be used as a transient signal feature.4.A method of steady-state signal feature and transient signal feature fusion based on SCCA algorithm is presented.A steady-state feature and transient feature fusion based on SCCA algorithm are presented.The experimental results show that the recognition rate of the steady-state feature(SIB)and transient signal polynomial fitting coefficient is 25% higher than that before fusion,15%,recognition rate of 98%.Compared with the PCA algorithm,the SCCA algorithm has an average recognition rate higher than about 10%,and the SCCA algorithm has a lower dimension after the feature fusion,and achieves a high recognition rate in the case of low feature dimension.In addition,SCCA algorithm has the advantages of low complexity.