Real-time EELS spectra classification using Principal Components Analysis (PCA) and Artificial Neural Networks (ANN)

Abstract

Recent developments in detector technology [1] for EELS and Energy Filtered TEM has made possible to obtain large number of spectra and energy filtered images during very short exposure times. This in turn opens the exciting possibility of studying time dependent processes in the electron microscope, during exposure to the electron beam as well as the study of different radiation sensitive samples which are being degraded during lengthily data recording. This kind of data recording generates a large amount of data and manual data analysis should be avoided in order to be able to fully benefit from the improved sensitivity and increased speed of these new detectors. Thus a fast, real-time data analysis system is highly desirable.A system for real-time data analysis (spectra classification) of data generated from such a detector has been simulated in a program based on the object oriented C++ framework ROOT [2][3].