Wavelet based feature extraction in near infrared spectra for compositional analysis of food
Abstract
Near infrared spectroscopy is a common method for analysis of food, soil and pharmaceutical products. New developments in sensor technology, like hyperspectral camera systems and mobile spectrometers, allow broad applications of spectroscopy with devices out of specialized laboratories. Therefore, it is necessary to develop robust algorithms for classification and regression, regardless of the device. The key to robust analysis lies in data preparation to get standardized spectral information from each device. Wavelet based feature extraction could be a possible method to compress spectral data to its material specific absorption information. A method for wavelet based feature extraction, which also reduces the influence from elastic scattering effects is proposed in this report.
