A Comprehensive Estimation Method for Kernel Function of Radar Signal Classifier

The current electromagnetism environment is fast changing and levity, the methods for evaluation Suppont vector machine (SVM) kernel functions which are used in radar signal recognition can not suit it. So kernel space separate, stability and parameter numbers were proposed in this paper to review the performance of kernel function, and a novel method for estimating kernel function was designed. By simulation, this novel method can estimate the performance of kernel function roundly, and can choose the best kernel function in different application demand for recognition.