Baseline correction for infrared spectra using adaptive smoothness parameter penalized least squares method

Abstract Baseline wander is a common problem in analysis with Fourier Transform Infrared Spectrometer (FTIR). And it is necessary to correct baseline drift for further quantitative and qualitative analysis. Several baseline correction algorithms based on penalized least squares have been proposed. However, these methods are usually used in noise-free or low-noise environments. In this paper, a novel algorithm named adaptive smoothness parameter penalized least squares was proposed. The smoothness parameters were set by user at first. Then, the smoothness parameter was updated iteratively according to the difference between the original spectrum and fitted baseline. When the iteration reaches the termination condition, the fitted baseline can be obtained. In the end of the paper, experimental results on simulated spectra and measured infrared spectra of methane were given. The simulated spectra results demonstrate that the proposed method has better performance than existing methods, especially when the spectra contain high noise. The results of infrared spectra confirm that the proposed method has good performance and can be applied to correct spectral baseline accurately.