Research on the Radio Transmitter Fingerprint Identification Algorithm
|School||Xi'an University of Electronic Science and Technology|
|Course||Communication and Information System|
|Keywords||Transmitter Fingerprint Individual Character Eigenvalue Decomposition Integral Bispectra Kernel Principal Component Analysis|
Since some subtle individual difference exists on the hardware of each transmitter,therefore there are some discriminative individual characteristics reflected on the transmitted signals, and transmitter fingerprint identification is a technology that focuses on seeking the source of the received signal based on these individual characteristics. So the transmitter tracking can be realized and a significant thread can be provided for the determination of communication network construction. Transmitter fingerprint identification developed from radar transmitter identification and it has becoming a key task to the communication reconnaissance in recent years.Based on the analysis to the individual characters, this paper makes a research on the transmitter fingerprint identification aiming at the different transmitters of the same type and the same working mode. First,With the basic theory and the mechanism, the traditional algorithms for transmitter identification are studied. Then, considering the deficiency that traditional LS-EVD algorithm has a low SNR estimation precision, a new method to construct the autocorrelation matrix is utilized, and thus an improved LS-EVD algorithm is proposed for the transmitter individual identification. The simulation results show that the improved algorithm has effectively decreased the error of the SNR estimation, greatly improved the property of transmitter identification. Finally, in order to cope with the deficiencies in the traditional algorithm based on integral bispectra which has lots of redundant features in the selected integral bispectra, and the redundant features may decrease the accuracy of classification greatly in low SNR, a method to reduce the feature dimension using the Kernel Principal Component Analysis Algorithm(KPCA) is utilized, and thus an improved transmitter identification algorithm based on Kernel Principal Component Analysis and square integral bispectra is presented. The simulation results indicate that the proposed algorithm has greatly reduced the feature dimension and improved the correct identification rate.