[EEMD de-noising adaptively in Raman spectroscopy].

Area of Science:

• Signal Processing
• Spectroscopy Analysis
• Data Denoising

Context:

• Traditional wavelet denoising requires empirical parameter tuning.
• Empirical Mode Decomposition (EMD) offers adaptive denoising but suffers from mode overlap in high frequencies.
• Raman spectra present challenges due to non-linearity, non-smoothness, and high-frequency noise.

Purpose:

• To evaluate Ensemble Empirical Mode Decomposition (EEMD) for adaptive denoising of Raman spectra.
• To compare EEMD's performance against Fast Fourier Transform (FFT), wavelet denoising, and EMD.
• To address mode overlap issues inherent in traditional signal decomposition methods.

Summary:

• EEMD successfully decomposed non-linear, non-smooth bean grease Raman spectra, yielding clear characteristic components by mitigating high-frequency mode overlap.
• Comparative analysis using signal-to-noise ratio, root mean square error, and correlation coefficient demonstrated EEMD's superior high-frequency denoising compared to FFT and EMD.
• Wavelet denoising showed comparable results to EEMD, but EEMD offers an adaptive denoising process without parameter setting.

Impact:

• EEMD provides an effective and adaptive alternative for denoising complex spectral data, particularly Raman spectroscopy.
• The study highlights EEMD's potential to improve the accuracy and reliability of spectral analysis in various scientific fields.
• Future research directions include exploring time-frequency spectrum analysis methods and establishing noise property criteria for Intrinsic Mode Functions (IMFs).