Measuring Emotional Arousal for Online Applications: Evaluation of Ultra-short Term Heart Rate Variability Measures

The objective of this paper is to examine the possibilities and limitations of heart rate variability (HRV) as an indicator of emotional arousal for mobile applications which require online biofeedback. In contrast to offline classification, feature extraction for online applications sets other requirements to the window size in which data is analyzed as the delay between a change of a person's arousal level and the reaction of an application should be as short as possible. For this purpose we compare various HRV features in order to evaluate how far window size can be decreased to enable online arousal recognition. Using data from a study where high and low arousal were induced in a game scenario, HRV features are analyzed for their discriminatory power depending on the window size using Fisher's discriminant analysis. Moreover, we use these features to train an SVM based classifier. Results indicate that for some features it is possible to use ultra-short term window sizes, i.e. window sizes shorter than the 5 minute window which has traditionally been used for short term HRV analysis.