Automated tracking of multiple body parts in video recordings of neonatal seizures

This paper presents an automated procedure for tracking multiple body parts in video recordings of neonatal seizures. This procedure detects motion by relying on optical flow computation and then tracks one or more body parts by employing predictive block matching. Predictive block matching estimates the displacement of a feature between two successive frames by minimizing an error function defined in terms of the feature intensities at these frames. Displacement estimation is followed by adaptive block matching based on Kalman filtering. The reliability of the proposed automated tracking procedure is illustrated by its application in the extraction of temporal motor activity signals from video recordings of neonatal seizures.