Study On Learning Algorithm For Unknown Radar Signal Reconnaissance And Processing

As the increasing diversity of the radar system, it makes the signal reconnaissance, processing, classification, identification more difficulty. The analysis of the pulse modulation characteristics will help improve the radar signal sorting, identification of reliability and speed.This paper introduced the theory of channel and digital receiver principles and advantages, and the form of the radar signal pulse description words, and common forms of modulation signals. Then the paper introduced some kinds of methods which analysis the modulation of the common signals: Instantaneous auto-correlation, Short-time Fourier transform, Wigner Distribution and wavelet analysis. An analysis on using range, advantages and disadvantages is made between them. A comparison on real-time, accuracy and noise immunity is made between them. And a method of classification and recognition of radar signal which is based on decision tree is given. A simulation is made, and a statistic of correct rate is given. Based on this, this paper gives a new method of radar signal identification which based on neural network. An analysis is made on the learning algorithm and the limitation of learning algorithm. And an improving method is given in the paper.Finally, a simulation is made in the method of neural network which is used in classification and identification of radiation sources. A statistic on correct rate is made between different training methods and different sample numbers. And in the last chapter a forward improvement is given.