Automated Analysis of Mixed Sample Raman Spectra Using Feedforward Neural Networks and One-Vs-All Decomposition
Batten College of Engineering & Technology
M.S. Engineering - Electrical & Computer Engineering
Interest in use of Raman spectrometers in many fields of analytical science has increased due to ability to nondestructively provide information about molecular structures and component materials of a mixed sample. Advancements in Raman spectrometer hardware has allowed for compact instruments to have deployment capabilities directly on interplanetary missions, flexible usage conditions requiring no sample collection/preparation, and no need for daylight radiation shielding. As the amount of science which can be collected from a Raman spectrometer in a given amount of time increases, a bottleneck will be created in data analysis which leaves a need for a faster method of spectral data classification. In this study, a framework to allow for fast automated analysis of mixed sample Raman spectral data is proposed and an implementation of this framework is tested. Analysis of mixed sample Raman spectra was achieved by implementing a model which decomposes an N-class multilabel problem into “N” single class detection problems. The model (consisting of multiple neural networks) was trained with pure sample data and was tasked with analyzing both real and theoretical mixed sample Raman data. Performance of the model is judged by its ability to detect component materials in real mixed sample data at the same level that it is able to in ideal mixed sample data (consisting of linear combinations of training data). The model’s structure, training and testing methodologies, and results will be presented.
Download Full Text (707 KB)
Atkinson, Alexander; Abedin, M. N.; and Eksayed-Ali, H. E., "Automated Analysis of Mixed Sample Raman Spectra Using Feedforward Neural Networks and One-Vs-All Decomposition" (2019). College of Engineering & Technology (Batten) Posters. 3.