Application of Radial Basis Function Neural Networks in Complicated Radar Signal Measurement and Sorting
Zhang Yongqiang, Sun Guozhi · 2007
An intelligent radar signal sorting system with a robust radial basis function (RBF) is presented in this paper. This system can automatically sort the random overlapped radar signal stream and separate the input pulse stream to individual radar pulse sequence. Because tradition Gaussian neural network uses Gauss function as its basis function and adopt gradient descending method to adjust parameters. So the tradition method is likely to produce some non-expectation in learning process. In order to solve the problem, the proposed RBF uses Log-Sigmoid function as its basis function, so it eliminates any risk of instabilities, and it has better learning properties and function approximation capabilities. This algorithm ameliorates the traditional algorithm and enhances the robust properties of learning process. For one thing, the method can adapt to the complicated electromagnetic environment demand due to its self-adapting capability. For another, it can overcome the difficulty that the data have too much noise due to the detection system faultiness. Simulation results demonstrate the obvious superiority of this algorithm.