Application of a Wavelet Packet Transform Based Radial Basis Function Neural Network to Analyze Overlapping Spectra

Abstract

This paper suggested a novel method based on wavelet packet transform and radial basis function neural network (WPTRBFN) in simultaneous spectrophotometric determination of Mn (II), Zn (II), Co (II) and Cd (II) combining wavelet packet thresholding denoising with radial basis neural network. Wavelet packet representations of signals provided a local time–frequency description, thus in the wavelet packet domain, the quality of the noise removal can be improved. Radial basis function network was applied for overcoming the convergence problem met in back propagation training and facilitating nonlinear calculation. In this case, by optimization, wavelet function, decomposition level, the numbers of hidden nodes and the width σ of RBFN for WPTRBFN method were selected as Symmlet 5, 1, 20 and 1.2 respectively. The relative standard errors of prediction (RSEP) for all components with WPTRBFN, RBFN and PLS were 7.4, 8.9 and 8.1 percent respectively. The proposed method has been successfully applied to analyze overlapping spectra and better than others.