Converging image processing and data mining for Raman spectroscopy analysis

Abstract

The aim of this paper is to show how numerical data analysis such as principal component analysis (PCA) and graphic processing unit (GPU)-based image processing techniques can be used for Raman spectra application. GPU-based parallel processing is used to extract representative peaks to identify the chemical origin of Raman spectrometry samples. We had also demonstrated how enhanced spectroscopy Raman analysis can be exploited over biomaterials samples using PCA that exploit each spectra feature correlated with properties (peaks) set in the basis on existing databases. We use BIORAD database in order to generate certain classes using Raman vibration peaks. The approach of PCA to data analysis gives promising results such as rapid clusters extraction. This classification can be combined to characterise Raman techniques.