Classifying head movements in video-recorded conversations based on movement velocity, acceleration and jerk

Abstract

This paper is about the automatic annotation of head movements in videos of face-to-face conversations. Manual annotation of gestures is resource consuming, and modelling gesture behaviours in different types of communicative settings requires many types of annotated data. Therefore, developing methods for automatic annotation is crucial. We present an approach where an SVM classifier learns to classify head movements based on measurements of velocity, acceleration, and the third derivative of position with respect to time, jerk. Consequently, annotations of head movements are added to new video data. The results of the automatic annotation are evaluated against manual annotations in the same data and show an accuracy of 73.47% with respect to these. The results also show that using jerk improves accuracy.