Quantitative analysis of terahertz signals using CWT-based spectrogram and Zernike image moments

Area of Science:

• Analytical Chemistry
• Spectroscopy
• Chemometrics

Background:

• Terahertz (THz) spectroscopy combined with multivariate calibration is widely used.
• Existing calibration models struggle with environmental changes, instrument variations, and substance alterations.
• This necessitates the development of more robust and adaptable calibration techniques.

Purpose of the Study:

• Introduce and evaluate a novel multivariate calibration method, CWT-ZM.
• Enhance the robustness and generalization capabilities of THz spectroscopic calibration.
• Address the limitations of current methods in real-world applications.

Main Methods:

• Utilize continuous wavelet transform (CWT) for time-frequency analysis of THz pulses.
• Employ Zernike moments (ZM) for invariant feature extraction from the 2D time-frequency plane.
• Combine CWT and ZM to create the CWT-ZM multivariate calibration method.

Main Results:

• The CWT-ZM method provides richer characteristic information in both time and frequency domains.
• Zernike moments offer invariance to affine transformations like peak shifting and baseline drift.
• The proposed approach yields a more robust regression system with improved generalization.

Conclusions:

• The CWT-ZM method demonstrates high capability in capturing compound concentrations and invariance to measurement variations.
• Experimental validation on pharmaceutical tablet datasets confirmed the approach's effectiveness across different THz instruments.
• CWT-ZM offers an extensible framework for diverse spectral analysis applications.