ABSTRACT This study proposed and validated a hybrid Raman spectral processing framework that couples an adaptive segment adaptive iteratively reweighted penalized least squares (ASairPLS) baseline algorithm with ensemble empirical mode decomposition (EEMD). The integrated design targets the simultaneous suppression of baseline drift and noise interference in raw Raman spectra, enabling a more reliable analysis of biologically relevant samples. Within the framework, ASairPLS partitions the spectrum and performs tailored baseline estimation for each segment, leading to significant improvements in background removal across complex spectral intervals. Additionally, EEMD decomposes the signal into multiple frequency components, facilitating the isolation of noise from essential spectral structures. Raman measurements on U‐87 MG cells confirmed the effectiveness of the combined strategy. Spectra collected with an integration time of 1 s after processing achieved a signal intensity and signal‐to‐noise ratio (SNR) comparable to those obtained from 5‐s acquisitions. For imaging tasks, the method enabled 0.1‐s Raman images to reach SNR and spatial distribution clarity equivalent to those of 0.5‐s images. Collectively, the proposed ASairPLS–EEMD framework offers a robust and efficient solution for analyzing Raman spectra under low SNR conditions and complex baseline fluctuations, thereby strengthening the practical utility of rapid Raman detection and imaging.
Wang et al. (Mon,) studied this question.