Detection and correction of non-calibrated spectral features in optical spectra based on a wavelet-transformation

摘要

Many spectroscopic montiroing techniques employ chemometric algorithms like principal component regression (PCR) for calibration and evalaution of optical spectra. Systems based on this method, however, suffer from unknown spectral features appearing after calibration, which may result in major errors. The detection of non-calibrated absorption lines is important for treating errors in chemical processes. For these two reasons, the detection and classification of non-calibrated absorption features is of great importance in on-line spectroscopy. A novel approach is proposed here. A wavelet representation of principal components and measured spectra is shown to be appropriate for detection of non-calibrated spectral features. The algorithm can also be applied in combination with partial least-squares (PLS).