[Simultaneous spectrophotometric determination of multicomponent mixtures by a soft thresholding wavelet-based radial...

Area of Science:

• Chemometrics
• Artificial Intelligence
• Signal Processing

Background:

• Quantitative analysis of multicomponent mixtures presents challenges in noise reduction and accurate regression.
• Existing methods like multivariate linear regression may lack the precision needed for complex samples.

Purpose of the Study:

• To develop and evaluate a novel Soft Thresholding Wavelet-based Radial Basis Function Neural Network (STWRBFN) method.
• To enhance the simultaneous quantitative analysis of multicomponent mixtures by improving noise removal and regression quality.

Main Methods:

• The STWRBFN method integrates wavelet soft thresholding with a radial basis function neural network.
• Key parameters including wavelet function, decomposition level (L), thresholding method, and RBFN spread parameter (sigma) were optimized.
• Two programs, PSTWRBFN and PRBFN, were developed for STWRBFN and RBFN calculations, respectively.

Main Results:

• The STWRBFN method demonstrated successful application in simultaneous quantitative analysis.
• Experimental results indicated that STWRBFN significantly outperformed the standard Radial Basis Function Neural Network (RBFN).
• Both STWRBFN and RBFN methods proved more powerful than classical multivariate linear regression.

Conclusions:

• The developed STWRBFN method offers a superior approach for the quantitative analysis of multicomponent mixtures.
• Combining wavelet soft thresholding with RBFN effectively addresses noise reduction and improves regression accuracy.
• STWRBFN presents a powerful alternative to conventional chemometric techniques for complex mixture analysis.