Neural network for AES spectra peak base background removal

One of the constituent elements in Auger electron spectroscopy (AES) spectra that make their automatic analysis next to impossible is the background. To overcome this obstacle and enable further processing, the background removal stage must provide the clear enough data for the techniques capable of automatically extracting the requested information from the AES spectra. It is also known that the background in AES spectra contains further information regarding the composition of the sample studied. Thus, although removed, this background should not be disregarded. In our work, the neural networks for data modeling or background function approximation were used. Neural networks were used for the determination of the general shape of the background in AES spectra. Neural networks are model-less approximators, meaning that they are capable of accomplishing the approximation tasks regardless of any prior knowledge on the nature of the modeled system. According to our analysis, three distinctive parts of the background in the spectra were proposed. One of them is the so-called peak base. The use of the neural network for the subtraction of the peak base is described further in this article. Copyright (c) 2012 John Wiley & Sons, Ltd.