Wavelet neural network based calibration curve fitting in the quantitative assay

In the present paper, the wavelet neural network (WNN) is adopted to fit the calibration curve during the quantitative assay of the gold immuno-chomatographic (GIC) strip of the serum marker-HCG (human chorionic gonadotrophin). Based on the wavelet transform theory, the structure and arithmetic of WNN are introduced. Considering how to improve the ability of extracting the signal feature of the network, the training energy function is modified using the information entropy of the hidden layer. The results indicate that the WNN has a better performance in the measure accuracy and reliability of fitting curve than the same size BP neural network.