Identification Of Individual HF Radio Transmitters

Radio individual identification is an important research topic of spectrum management, communications reconnaissance and countermeasures. By extracting the subtle features of the radio signals, pattern recognition method can effectively identify individual radio. Radio individual identification in the field of military and civilian fields has a wide range of applications. For military applications, radio individual identification can grasp and monitor. For civilian areas, radio individual identification can be used in radio spectrum management, security-aware of the band, the detection signal interference, frequency conflicts and other acts of unlawful interference.This paper studies the steady-state characteristics of short-wave radio communications, and analyzes the spurious characteristics and the higher-order statistics features of steady signals, and using the measured data to extract them. PCA is proposed based on a single feature dimension reduction of SVM classification algorithm, which is suitable for simulation. In order to meet the needs of real shortwave system, we further proposed an algorithm based on the estimated SVM combination K neighbor classifier. Practical employment shows that our algorithm is effective.The main contributions of this paper is summarized as follow.(1) Indicating that the fine features of shortwave communication signals are generated by the production process and component performance and other random factors, these random factors lead to the crystal oscillator frequency difference between the nominal value, and result in a carrier frequency deviation and modulation deviation parameters.(2) Stray characteristic features and higher-order statistics shortwave radio signals were analyzed to extract the information dimension of higher-order spurious features J features and fractal dimension, LZC, box dimension and so on. Experiments show that higher-order J and LZC and Kx provide the best performances. We should reduce the dimension before we use the SIB.(3) Propose a SVM classifier algorithm based on PCA dimensionality reduction. The algorithm uses PCA to reduce the dimension of the SIB, and the dimension of the SIB reduces from 80 to 8. Then, after dimensionality-reduced SIB and other features are combined into a set of features as the input to the SVM classifier. Finally, recognition experiments illustrate the characteristics of post-processed have good classification performance, and more features for radio subtle differences distinguish the better.(4) In order to meet the needs of short-wave radio individual identification, we propose a SVM classification algorithm combined K neighbor estimate. The algorithm uses the advanced combined classification algorithms in pattern recognition, and meets the recognition rate.