Neonatal events recognition using LBP descriptor and wavelet thresholding technique

This paper presents an approach for preterm events recognition using local Binary pattern LBP and wavelet thresholding. Unlike conventional approaches which adopt contact measurement of vital signal and develop models for monitoring the preterm in neonatal intensive care unit (NICU), our approach describes non-contact monitoring vital signal based on thermal video as a new premature pain infant profile (PPIP). An interest clinical vital signal is face temperature of neonate. In this study, we collaborate with premature Infant events `pain and normal states´ detection and recognition during daily care monitoring. The proposed approach is composed of a thermal video analysis component and an activity recognition component. A video analysis component contains features descriptor using histogram of Local Binary Pattern LBP in the first experiment, and applying wavelet thresholding technique on approximation wave coefficients of histogram feature in the second experiment. Activity recognition contains classification of premature events using classifiers. We present recognition results by considering a variety of classifiers.