LAUSR

The spectrum acquired on the optical instrument usually contains the pure spectrum and undesirable components such as baseline and random noise. However, the intensity of the baseline, which seriously submerges the spectrum, is the primary limitation of spectral applications. Thus, baseline correction has become one of the most significant challenges for spectral applications. In this paper, we propose a doubly reweighted penalized least squares method to estimate the baseline. This method utilizes the first-order derivative of the original spectrum and established spectrum as a constraint of similarity. Meanwhile, the doubly reweighted strategy achieves a better effort. Considering the drawbacks of the weighting rules for the adaptive iteratively reweighted penalized least squares method, we adapt a boosted weighting rule based on the softsign function, which performs well when the spectrum contains high noise. The simulated results confirm that the proposed method yields better outcomes. The proposed method can be applied to Raman and near-infrared spectra as well, and the result shows that it can estimate various kinds of baselines effectively.