Overlapping Peaks Resolve Based On RBF Neural Network

With the continuous advances in the development of computer technology, instrumentation, fiber optic technology and metrology methods, in many areas, such as farm products, food manufacturing, chemical, pharmaceutical manufacturing, tobacco industry’s wide range of applications such as near-infrared spectroscopy.In spectral analysis studies often encounter overlapping spectral, the study of overlapping peaks is a important aspects of the spectrum of applications. A good analysis method lay a good foundation for further study, overlapping peaks of peak method is an important part of the study.Data pre-processing and analysis of overlapping peaks is the main research methods to explore the direction of effective data processing and analysis methods can effectively help us to analyze the spectral composition further research on substance. Under conditions of limited experimental conditions will be useful to contain all kinds of noise and interfering signals, if appropriate data pre-processing and analysis, decomposition of overlapping peaks will reduce the reliability and accuracy, and thus can not be able to use to give the desired valuable information. Data preprocessing methods can be selected according to the type of modeling and the actual situation. Partial least squares can put data decomposition and regression analysis together, suitable for nonlinear system modeling data, but the modeling process is complex and not easy to operate. For serious overlapping peaks valuable signal is weak, noisy situations, we can use artificial neural network modeling method.Radial Basis Function (RBF) neural network is an effective feed-forward neural network, which has the best approximation and the global optimum performance while fast method is simple, there is no local optimization problem, based on these RBF network has been a lot of advantages of the application of nonlinear time series forecasting issues. Gaussian fitting is a commonly used method to resolve overlapping peaks, you can use this method to get the naked eye can not distinguish between overlapping peaks, get relevant data from overlapping peaks.This paper describes the RBF neural network Gaussian fitting method, including single peak, multimodal analysis and analysis of overlapping peaks. Through the use of modeling and program analysis, in order to draw an intuitive graphical view and degree of matching of the original peak shape, get relevant spectral data.