Study On Fingerprint Features Of Frequency Hopping Spread Spectrum Signals

Frequency hopping spread spectrum (FH-SS) is one of spread spectrum technology which has advantages of high noise immunity and high interception resistance. With the rapid development of microelectronics and digital signal processing, the former problems existed in frequency synthesizer and frequency hop synchronization have been resolved. Now, FH-SS has not only adopted broadly in military telecommunication, but also has good prospects in application in civilian use. Therefore, it has great practical significance in research on the features of FH-SS.In order to meet the need of electronic war, parameter estimation for frequency-hopping (FH) signals has become the focus of the research at home and abroad. In this situation, many algorithms of parameter estimation have been presented. However, most of these algorithms did not considerate the own features of the real signals and some of them need to know some parameter of the signals. All of these have made these algorithms become limited in non-cooperative communication environment.In this paper, the higher order statistic features and wavelet features of FH-SS signals in non-cooperative communication environment have been studied, including the first generation wavelet and the second generation wavelet transformation features. With construction of different kernel functions using in SVM which has been introduced to use as the separator, we can realize the separation of different FH-SS signals. The kernel functions include polyspectral kernel and first generation wavelet kernel, and put forward a new algorithm of second generation wavelet kernel function. Finally, the measured radio data has been used to verify the feasibility of these kernel functions through the computer simulation. The primary work included in this paper can be described as followed.1) Higher order spectrum features of FH-SS signals had been studied using polyspectral kernel function. This method can reduce the numbers of the support vectors effectively through mapping the lower order inseparable samples to higher spaces. Moreover, it can improve the identification rate as well. The experiment shows that better and stable classification rate can be achieved.2) Wavelet transmission features of FH-SS were studied in this paper. We used the SVM as the separator to classify different measured FH-SS signals. The first generation wavelet kernel function was used in the SVM. The simulation results demonstrate that this algorithm can significantly improve the classification rate which meets the requirement of the practical application, however, the efficiency was still to be improved.3) In order to improve the efficiency of classification, a new algorithm was put forward, that is the second generation wavelet kernel function. Through the simulation experiment, the performance of this kernel function has been discussed. A conclusion has been made that the second generation wavelet kernel function can not only obtain a good classification rate but also improve the classification efficiency.