Protein content estimation in food products using neural networks

Abstract

In the food production industry, estimation of protein content in various forms of products is a continual process where the estimated values are required for documentation as well as testing purposes. Chromatographs and infrared spectrometers are used to physically obtain the protein spectra from the food samples. The spectra provides some measures of protein contents of the samples. The use of faster on-line estimation programs with sufficient accuracy is desirable. A recent study indicates that the use of neural networks for the task of protein estimation is highly feasible. We followed upon the idea and modeled a few type of neural networks. These network types include back-propagation networks (BPNs), genetic reinforcement networks (GRNs), and radial basis function networks (CRBFNs). We found that the tested models are usable for the estimation purposes. In this article, we present our modeling and test results. © 1999 John Wiley & Sons, Inc. Lab Robotics and Automation 11: 151–155, 1999