A method for identification of unknown-samples from simultaneous multi-spectrum measurement data employing Laser Induced Breakdown Spectroscopy (LIBS)

Abstract

This study proposes a new identification method for unknown-samples from multi-spectrum measurement data by LIBS. So far, in order to identify unknown chemical compounds, NIST LIBS Database -Atomic Spectra Database- are used. We introduce machine learning methods (e.g., least squares with penalty terms, NMF, decision trees, etc.) to not only identify unknown chemical compounds, but to automate a series of processes from preprocessing of measured data to classification and estimation of it.