FTIR spectral subtraction based on asymmetric least squares

Based on asymmetric least squares, a new method for infrared absorbance spectral subtraction is proposed. In order to reflect the scaling relation between the sample spectrum and scaling spectrum, the proposed method uses scaling vector rather than scaling factor which is usually used in the traditional spectral subtraction model. As independent of reference peak, the method can deal with most of subtraction cases of binary mixtures. For multiple system with unknown scaling spectrum, ei-genspectra method is introduced to calculate the accurate scaling spectra. Experimental results show that difference spectra can match the real spectra well. Besides, in contrast to the results of traditional method, difference spectra obtained by this method is more similar with the real spectra. The prediction results based on partial least squares (PLS) model demonstrate that the performance of difference spectra is better than that of sample spectra in characteristic regions.