Stochastic resonance-based Raman spectroscopy denoising

Raman spectroscopy is a non-destructive technique that analyzes the vibrational modes and other properties of molecular systems by measuring scattered light from a laser. However, due to the short exposure time and low power of the excitation laser, Raman signals are often very weak, sometimes even weaker than the noise, making them prone to being overwhelmed by noise. This reduces sensitivity and classification accuracy, affecting practical applications. Currently, traditional denoising methods face several challenges, primarily because their effectiveness heavily depends on manual parameter tuning, which is not beginner-friendly and adds complexity to using these methods. This study proposes an Adaptive Bistable Stochastic Resonance (ABSR) system, which enhances signals by utilizing noise energy and adjusts system parameters through a Particle Swarm Optimization (PSO) algorithm, eliminating the need for manual parameter tuning to achieve optimal signal enhancement. In the experimental section, the denoising performance of the ABSR algorithm was systematically validated using simulated Raman spectra. The experimental results demonstrate that, compared to traditional methods such as Savitzky-Golay (SG) filtering, Gaussian filtering, Soft and Hard Threshold Wavelet Transform (SHTWT), Adaptive Savitzky-Golay (ASG), and Stein's Unbiased Risk Estimate Wavelet Transform (SUREWT), the ABSR algorithm exhibits significant advantages in denoising effectiveness. Specifically, ABSR is more effective in preserving the detailed features of spectral signals while demonstrating superior performance in noise suppression. Besides, using Raman spectra from diabetic patients, ABSR showed significant improvements in SNR and RMSE, and performed better in classification algorithms like SVM, Random Forest, and Decision Trees. The ABSR method effectively enhances Raman spectral resolution, reduces laser exposure, and is simple to use, making it valuable for beginners in Raman spectroscopy research.