No abstract

The spectral analysis depends heavily on unwanted signals, such as the fluorescent background from the samples or other interfering components. A number of mathematical algorithms have been proposed to remove the background of Raman spectra. However, these methods require the selection of appropriate parameters to correct the of Raman spectra baseline. In this paper, we propose a method of adaptive noise model based on iteratively reweighted penalized least squares (ANM‐IRPLS) for Raman spectrum baseline correction. The algorithm was applied to various artificial spectra containing real forms of baselines and characteristic Raman peaks and then to the spectra of real drug samples with fluorescence obtained on a device equipped with a 532‐nm laser with a resolution of 15 cm−1. The modeling results showed that the proposed ANM‐IRPLS baseline correction method allows for better results in background removal than the airPLS. For real Raman spectra processed by the ANM‐IRPLS method, it is shown that the algorithm handles a complex background well, while maintaining the characteristic Raman signal features, such as a wide water peak for aqueous solutions.

Development of drug quality control methods using non‐invasive methods, in particular Raman spectroscopy, is an important task in the pharmaceutical industry. This paper presents a new approach to the analysis of drug spectra obtained with portable Raman spectrometers. It represents a normalized relative error distribution analysis (nReDA). The proposed method uses the Pearson residuals distribution statistics and, compared with the correlation method, allows increasing the screening sensitivity to foreign components that are not contained in the original sample, with the same resistance to changes in the noise distribution and the peak ratios. On the basis of the statistical data, the nReDA in some way is similar to the techniques such as principal component analysis, however, it requires a considerably less spectra to build a representative model and is suitable for comparing Raman spectra, having a noise other than the Gaussian one. The paper shows the prospects of the algorithm for creating a spectral library and solving the problem of transferring Raman libraries to a large number of spectrometers of the same type.