[S-MP23] Physics and Chemistry of Minerals

[SMP23-P09] Application of MATLAB for Mineral Classification in Raman Spectroscopy

*Hsu I Huang¹, Huei-Fen Chen¹, Yi-Ling Huang¹ (1.National Taiwan Ocean University)

Keywords:Mineral classification, Raman spectroscopy, Matlab, Baseline processing

In today's material identification industry, various instrumental analysis methods are combined to accurately identify unknown minerals and obtain qualitative or quantitative results. Raman spectroscopy, with its non-destructive analysis characteristics, has become the preferred method in the jewelry appraisal industry. The comparison software used in Raman spectrometers is mostly provided by instrument developers, so are the databases. Although instrument developers maintain these databases, they lack support from academic papers, making it difficult for academic researchers to use them for publication purposes. Therefore, this study introduces the open-source database provided by the RRUFF project at the University of Arizona in the United States to establish a credible comparison source for mineral data processing and identification software development for academic work. The development platform used is MATLAB, a commercial mathematical software from The MathWorks company, which excels in numerical computation and provides important functions such as signal processing, data analysis, and visualization. For user convenience, the software operation interface is designed in a graphical user interface manner, integrating systems such as database construction, sample preprocessing, data comparison, and result plotting. In the sample processing workflow, samples are imported into the software through the database system and undergo background processing using an improved Savitzky-Golay filtering function to remove spectral baselines. Smoothing is then carried out through a denoising filter—low-pass filter—followed by finding Raman shifts of peak values using a local maximum algorithm. During sample comparison, the sum of differences in peak values is sorted to display the data in the database most similar to the sample. The result shows that the software design is compatible with the data from most open-source databases, and in terms of mineral comparison functionality, it can classify single mineral data and remove peak values of individual minerals from multi-mineral phase data using the software.