Identification of electromagnetic radiation source with support vector machines
Dan Shi, Junjian Bi, Chao Li, Zhiliang Tan, Hongbo Wang, Gao Yougang · 2015
A method for electromagnetic radiation source identification is proposed. The spatial characteristic of a radiation source is taken as the unique parameter for support vector machines (SVMs) to identify. First, the location of radiation source is determined by the triangulation method, and then its spatial characteristic is collected by a band receiver array with simulation, which removes the limit of absolute similarity between test data and training data. The 3D data are converted into a 1D vector with subscripts as inputs for SVMs, which are trained by the inputs to identify radiation source types intelligently. The identification time needs a few seconds, much faster than artificial neural networks (ANNs). The influence of parameters (e.g., noise from ambient environment, data collection method, scaling method for inputs, and parameters of SVMs) is discussed. The proposed method has good performance in noisy environment and the identification accuracy is 76.57 %, even though the signal to noise ratio decreases to 10 dB.