On nonhomogeneous activity estimation in Gamma spectrometry using sparse signal representation

One of the main objectives of spectrometry is to estimate the activity of a radioactive source by means of the analysis of the photon energy received by a detector. However, most methods of activity estimation rely on the fact that it is constant over the time of experiment. We provide in this paper an algorithm based on a sparse representation of the signal, which allows to estimate the counting rate activity efficiently even in the non-homogeneous case. This algorithm can be summarized as follows: a sparse regression is applied on the incoming signal, giving an estimation of the arrival time of each photon. These estimates are arfterwards plugged into a standard nonparametric estimator of the activity. We investigate the efficiency on the proposed approach on simulation and real datasets.