Denoising method for Raman spectra with low signal-to-noise ratio based on feature extraction

Area of Science:

• Spectroscopy
• Analytical Chemistry
• Biophysics

Background:

• Raman spectroscopy is a non-destructive technique for molecular analysis.
• Weak Raman signals and low signal-to-noise ratios (SNR) pose significant challenges for conventional methods.
• Precise extraction of Raman peaks from noisy spectra is crucial for accurate molecular identification.

Purpose of the Study:

• To develop and validate a novel denoising method for Raman spectra with low SNR.
• To enhance the extraction of characteristic Raman peaks from complex and noisy samples.
• To demonstrate the method's effectiveness and practicality for analyzing biological samples.

Main Methods:

• Proposed a denoising method based on feature extraction and Hilbert Vibration Decomposition (HVD).
• Decomposed Raman spectra into components to locate and compensate for peaks.
• Reconstructed Raman peaks using Gaussian signals based on extracted peak features (position, height, FWHM).

Main Results:

• Significantly improved SNR from 3.5316 to 130.6386 and reduced MSE from 213.8635 to 14.0404 in simulations.
• Successfully extracted melamine characteristic peaks using a low excitation laser (10 mW), matching results from a high laser power (150 mW).
• Accurately identified numerous characteristic peaks in biological samples like rice leaves, including carotene, proteins, nucleic acids, and cellulose.

Conclusions:

• The feature extraction-based denoising method effectively extracts Raman peaks from low SNR spectra, even when submerged in noise.
• The method is practical due to its simplicity, few parameters, and ease of implementation.
• This technique offers a robust solution for analyzing complex biological samples with weak Raman signals.