Analysis Of Weak Target Detection Method Based On Least-squares Support Vector Machine In The Chaotic Background

Detection and process of weak target signal is an important direction of the signal processing. Target signal is easy to be covered by noise when the detected signal is weak, so that the detection is rather difficult and the main factor that effects the detection of the weak signal is the interference of the noise. To extract target signal effectively from the chaotic noise has great impact on the signal detection.This article utilizes the generalized window function and phase space reconstruction theory to estimate the basic parameters needed when signal is reconstructed, establishes the chaotic predicting model by the Least Squares Support Vector Machine (LS-SVM) and detects the weak target signal such as weak transient signal and periodic signal, from the prediction errors in the chaotic background. Specific research points as following:Reconstructing the original dynamical system in the multidimensional space is the most important basis of forecasting the chaotic time series. Hence, the time delay and the embedded dimension have a great influence on the reconstruction of the phase space. This paper studies the phase space reconstruction theory of the complex nonlinear system based on Takens Theorem, employs the improved autocorrelation to determine the optimal time delay and embedded dimension of the system, approximates the original dynamical system so that the characters of the primary system can be researched to improve the accuracy in signal analysis.LS-SVM is employed to build the detecting model in chaotic background and simulating experiment is made for the weak target signal like weak transient and periodic one within the Lorenz system to demonstrate that this model has relatively smaller error and can find target signal effectively with low effect on the target by the noise. Compared to the traditional methods, such as RBF neural network, the prediction model utilizing the generalized time window function and LS-SVM lowers the detecting threshold and enhances the precision. When the signal-to-noise ratio is-87.41dB in the chaotic noise background, the method in this paper reduces the root mean square error to0.000036123by nearly two orders of magnitude, while the traditional SVM can only reach to0.049under the condition of-54.60dB.Chen’s system is used in analyzing and verifying the results, which is acquired on the basis of researching the Lorenz system. It shows that the method enhances the general applicability which is able to approximate the actual physical process and detects the weak target signal effectively. What’s more, the model set up by the method in the paper has general use to some extent.This article utilizes Chen’s chaotic system to prove the method that it enhances the general applicability which is able to approximate the actual physical process so that the prediction and detection perfection can be reached.