Multi-medial stress assessment

A research project on stress assessment is running at Delft University of Technology since 1992. One of the aims of the project is to develop an instrument for automated stress assessment. The underlying system is based on the analysis of facial expressions, voice analysis and the analysis of physiological signals such as heart rate and blood pressure. Analyses of these multi-medial data takes place in parallel and are based on artificial intelligence technology. In each of the parallel subsystems, corresponding to sensor, image and sound data, the functionality is split-up into a number of layers: filtering and reduction layer, preprocessing layer, processor layer, application layer and output layer. A central interpreter, resulting in an overall stress measure, combines the results of the analysis. In this paper the stress assessment is used to monitor vigilance levels of car drivers with a focus on voice analysis.