Specific Emitter Identification Using Regression Analysis between Individual Features and Physical Parameters

In this paper, a semi-physical simulation platform and ADS are used to acquire the signals of three types of specific emitters, then high-order spectral analysis and variational modal decomposition are used to extract features of the signals. The phase noise of the oscillator and the bias voltage of the power amplifier are used as independent variables to study their influence on the features, based on which the correlation analysis is carried out. Regression fitting is performed on the variables with significant correlation to obtain a regression function, then a feature-weighted support vector machine is constructed for classification. The results show that the accuracy of the proposed identification algorithm using regression analysis is more than 10 percent higher than that of the single-kernel support vector machine under the same signal-to-noise ratio.