Performance metrics for energy spectrum estimation

Abstract

Nuclear spectroscopy frequently involves estimating the distribution of photon energies (energy spectrum). The evaluation of spectral estimates within the literature are often subjective. Efforts to quantify the quality of the estimation generally utilize a measure based on the integrated square of error. The ability to characterize low concentration peaks, and pulse pile-up distortion are important in numerous spectroscopic applications. The integrated square error is shown to provide relatively poor indication of both moderate pile-up and when low concentration peaks are present. Alternative measures for quantifying the quality of estimated spectra are suggested.