Adaptive Savitzky–Golay Filter for Denoising Gas Mixture Absorption Spectra

Abstract

Quantitative analysis of the gas mixture absorption spectra is complicated by noise. The parameters of standard filters are related to the entire analyzed spectral range. This means that the filter parameters being optimal for strong absorption lines are not optimal for weak absorption lines and vice versa. An approach to create adaptive filter for denoising experimental spectra based on the combination of a windowed version of a standard filter with the independent component analysis is suggested and implemented with the Savitzky–Golay filter as an example. The numerical simulation was carried out at normal conditions for the absorption spectra of the model of mid-latitude summer atmosphere in the 100–1000 GHz spectral range. The efficiency of the suggested adaptive and the standard versions of Savitzky–Golay filter was compared using a quantitative criterion of the proximity between two spectral curves. Experimental validation of efficiency of the suggested adaptive Savitzky–Golay filter was conducted on the example of 200 ppm SO₂ and 10 000 ppm H₂O gas mixture. The SO₂ concentration was evaluated using multivariate curve resolution method. The relative error in the concentration retrieved after noise reduction by this filter was 3.7 times less compared to the standard Savitzky–Golay filter. Thus, the suggested adaptive Savitsky–Golay filter makes it possible to increase the efficiency of denoising experimental spectral data.