Detection of Weak Signal in Chaotic Clutter Using Advanced LS-SVM Regression

In this study, detection of small target in chaotic clutter with unknown dynamics is presented. We achieve this in four steps: (i) by using db3 wavelet decomposition of the signals, (ii) using Takens delay embedding theorem and least-squares support vector machine (LS-SVM) prediction, including increase the symmetric constraint and improve the kernel function, (iii) wavelet reconstruction, (iv) separation the weak signals from the prediction error. Efficiency of the new approach is evaluated by computing the root mean square error (RMSE) and signal-noise-radio (SNR) of the estimation. Lorenz attractor and the data from the McMaster IPIX radar sea clutter database will be used in the simulation. It is demonstrated in the simulation that compared with conventional RBF neural network LS-SVM regression prediction method; this approach has stronger generalization ability and better accuracy.