Score-moment combined linear discrimination analysis (SMC-LDA) as an improved discrimination method

Area of Science:

• Chemometrics
• Spectroscopy
• Data Analysis

Background:

• Principal Component Analysis (PCA) scores are commonly used for spectral feature representation.
• Spectral moments offer an alternative feature representation method in spectroscopy.
• Linear Discriminant Analysis (LDA) is a standard technique for classification tasks.

Purpose of the Study:

• To develop and evaluate a new discrimination method, Score-Moment Combined Linear Discriminant Analysis (SMC-LDA).
• To assess the effectiveness of integrating PCA scores and spectral moments for improved discrimination.
• To compare SMC-LDA performance against conventional PCA-LDA and moment-based LDA.

Main Methods:

• Developed SMC-LDA by combining PCA scores and spectral moments as inputs for LDA.
• Evaluated three approaches: PCA-LDA, moment-based LDA, and SMC-LDA.
• Utilized three spectroscopic datasets: IR spectra of stomach tissue, NIR spectra of oils, and Raman spectra of ginseng.

Main Results:

• SMC-LDA consistently achieved the best discrimination results across all three tested datasets.
• The combination of PCA scores and spectral moments provided more diversified and descriptive information than either feature alone.
• This suggests that score and moment capture complementary spectral information.

Conclusions:

• SMC-LDA offers a superior approach for spectroscopic data discrimination compared to conventional methods.
• Integrating spectral moments alongside PCA scores significantly enhances classification performance.
• The findings highlight the value of combining diverse spectral feature representations for robust analysis.