Knowledge-Driven Interpretation of ESCA Spectra

Abstract

A system for automatically interpreting raw spectral data of an ESCA (Electron Spectroscopy for Chemical Analysis) is described. This system can remove background noise from the spectrum, identify each spectral peak, assign each peak to a corresponding atomic orbital and can obtain the chemical composition of a sample. A knowledge-driven method enables the system to identify very small peaks and rider peaks, and to obtain the most reliable assignment to each peak on the basis of specific chemical knowledge. Utilization of knowledge in low Level processing is our main interest, while higher-level logical inference is emphasized in other knowledge-based systems, for example DENDRAL and MYCIN.