Development of a context model based on video analysis

Abstract

This paper reports on the analysis of video footage of a human-robot study in public place regarding context factors that significantly influence the interaction. Therefore, a coding scheme was developed which was later used to analyze the video footage of the human-robot study. To ensure the validity of the coding, the video footage was coded independently by two students and afterwards the Cohen's Kappa was calculated to ensure the intercoder reliability. This calculation served as a basis for the creation of the first context model which shows factors that influence the interaction. This approach could show that it is possible to extract valid context factors and to create a context model based on video annotation. These factors should then be further tested in lab-based studies to get a better understanding of how they affect the human-robot interaction.