Resampling Method to Compensate the Predicted Error Induced by Spectral Shift for Partial Least Squares (PLS) Models in Laser Raman Spectroscopy
H. Bian, J. Wang, Y. S. Yu, X. Y. Wang and J. Gao
We propose herein an algorithm to resample the Raman spectra which aims to correct the predicted error induced by spectral shift. The same set of training data for both human and nonhuman blood Raman spectra was applied to build the model using partial least squares (PLS) method. Several kinds of Raman spectra of the blood samples originating from human, goose, monkey, chicken, sheep and duck without spectral shift were used to validate the model. The Raman spectra of the same blood samples with a shift of 9.18 cm-1 were used to illustrate the influence of the spectral shift to the model. To demonstrate the effectiveness of the algorithm proposed, Raman spectra with spectral shift were resampled. The results show that the resampling algorithm can improve both the stability and accuracy of the model and reduce the impact of the spectral shift, which is promising for quantitative or qualitative analysis in Raman spectroscopy for complex samples.
Keywords: Diode laser, Raman spectroscopy, partial least squares (PLS), spectral shift, resampling algorithm, blood discrimination