Pain intensity estimation by a self-taught selection of histograms of topographical features

Pain assessment through observational pain scales is necessary for special categories of patients such as neonates, patients with dementia, and critically ill patients. The recently introduced Prkachin-Solomon score allows pain assessment directly from facial images opening the path for multiple assistive applications. In this paper, we proposed a system built upon the Histograms of Topographical (HoT) features, which are a generalization of the topographical primal sketch, for the description of the face parts contributing to the mentioned score. We further propose a semi-supervised, clustering oriented self-taught learning procedure developed on the Cohn-Kanade emotion oriented database by adapting the spectral regression. To make use of inter-frame pain correlation we introduce a machine learning based temporal filtering. We use this procedure to improve the discrimination between different pain intensity levels and the generalization with respect to the monitored persons, while testing on the UNBC McMaster Shoulder Pain database. We introduce the Histogram of Topographical (HoT) features to address the variability in face images.We propose a semi-supervised, clustering-oriented, self-taught learning procedure.We propose a machine learning based, temporal filtering to increase the overall accuracy.A system for face dynamic analysis that applied to pain intensity estimation leads to qualitative results.