Research on heterogeneous noise suppression and peak parameter inversion in Raman spectroscopy under a three-stage synergistic framework

Le résumé fourni par la source

The precise inversion of weak signals in Raman spectroscopy is challenged by multiple, spectrally distinct noise sources, including high-frequency random noise, ultra-low-frequency baseline noise, and background fluctuations proximate to the signal, which conventional signal processing techniques struggle to address simultaneously. To overcome these limitations, this study proposes a synergistic DWT-AIRPLS-CWT (Discrete Wavelet Transform-Adaptive Iteratively Reweighted Penalized Least Squares-Continuous Wavelet Transform) strategy that enables the robust recovery of weak Raman signals through a refined, staged, and targeted suppression of noise. The strategy first employs the Discrete Wavelet Transform (DWT), leveraging its multi-resolution decomposition to concentrate high-frequency noise energy into detail coefficients that are then suppressed via soft-thresholding, thus maximally preserving the core spectral information in the approximation coefficients. Subsequently, the non-parametric Adaptive Iteratively Reweighted Penalized Least Squares (AIRPLS) method adaptively removes the persistent ultra-low-frequency baseline noise, a critical preprocessing step to isolate the Raman signal and its proximate fluctuations. Finally, the Continuous Wavelet Transform(CWT), using a Mexican Hat wavelet matched to the peak shape, is applied for fine-grained signal identification. CWT ridge analysis robustly distinguishes the long, strong ridges of authentic Raman peaks from the short, weak ridges of artifacts and directly provides high-quality initial parameters for peak fitting, with positions determined by ridge locations and widths estimated from scales of maximum response. These high-quality initial parameters significantly enhance the convergence and accuracy of the subsequent non-linear least squares fitting based on the Voigt function, ultimately achieving a high-precision inversion of extremely weak Raman signals.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.