Neural network pattern recognition by means of differential absorption Mueller matrix spectroscopy

Area of Science:

• Spectroscopy
• Biophysics
• Computational Chemistry

Background:

• Polarized light scattering provides unique signatures for organic molecules.
• Mueller matrix measurements capture polarization changes, offering rich data for analysis.
• Developing rapid, accurate detection methods for organic compounds is crucial in various scientific fields.

Purpose of the Study:

• To develop an artificial neural network (ANN) system for identifying solid organic matter.
• To utilize polarized light scattering signatures, specifically Mueller matrices, for analyte detection.
• To create a filter function based on ANN training for discerning specific organic compounds.

Main Methods:

• Constructed ANN systems trained using backward-error propagation and adaptive gradient descent.
• Analyzed polarized light scattering signatures in the form of Mueller matrices.
• Developed a weight matrix filter derived from ANN training to identify analytes.

Main Results:

• The trained ANN system successfully recognized patterns in Mueller matrices corresponding to specific analytes.
• A weight matrix filter was generated, capable of discerning analytes based on their unique polarization signatures.
• The filter function demonstrated potential for implementation in future spectroscopic instruments.

Conclusions:

• ANNs are effective tools for analyzing complex polarized light scattering data.
• The developed method offers a novel approach for the label-free detection of organic matter.
• The filter function has potential applications in advanced spectroscopic analysis and material science.