Quaternion-based Parallel Feature Extraction: Extending the Horizon of Quantitative Analysis using TLC-SERS Sensing

Area of Science:

• Analytical Chemistry
• Spectroscopy
• Chemometrics

Background:

• Quantitative analysis using thin-layer chromatography-surface-enhanced Raman scattering (TLC-SERS) faces challenges due to process variations and random substrate properties.
• Traditional chemometric methods for TLC-SERS analysis often overlook spatial distribution and correlations among sampling points, limiting accuracy.

Purpose of the Study:

• To develop a novel parallel feature extraction and fusion method for quantitative TLC-SERS analysis.
• To address the limitations of traditional methods by incorporating spatial information and signal correlations.

Main Methods:

• Proposed a parallel feature extraction and fusion method based on quaternion signal processing.
• Utilized quaternion principal component analysis (QPCA) for feature extraction and fusion.
• Employed support vector regression (SVR) for quantitative modeling of melamine in milk samples.

Main Results:

• The proposed quaternion-based method significantly improved the accuracy of melamine quantification in milk.
• Achieved low quantization errors of 7% and 2% at 20 ppm and 105 ppm melamine concentrations, respectively.
• Validation testing demonstrated consistently smaller measurement errors and variance compared to traditional TLC-SERS methods.

Conclusions:

• The quaternion signal processing approach is effective for quantitative sensing applications using TLC-SERS.
• This method offers a robust solution for overcoming variations and improving accuracy in complex analytical techniques.
• The study highlights the potential of QPСA for advancing quantitative analysis in chemical sensing.