Options
2014
Conference Paper
Titel
Automatic Video Analysis for Annotation of Human Body Motion in Humanities Research
Abstract
The analysis of multi-modal audio-visual data is the very first step to perform research on gestural behaviour in a variety of disciplines such as psycho-linguistic, psychology, behaviour analysis or sign language. The annotation of human gestures and motion of hands is a very time consuming process and requests a lot of effort in terms of personal. Due to the large amount of available data in existing corpora, much video material has not been annotated and even not touched, i.e. valuable material cannot be exploited for research. Thanks to modern video processing algorithms some of the most time consuming annotation tasks can be performed automatically. In this paper, we present a video analysis tool that is specifically designed to perform automatic analysis and annotation of video material for the above mentioned research domains. The presented algorithm provides a large variety of annotations required for behaviour analysis without any user interaction in a fully automatic way. The proposed video analysis tool is currently designed to provide annotations according to the NEUROGES coding system for gestural behaviour, but it can provide also other means of annotations for other coding schemes.