Hier finden Sie wissenschaftliche Publikationen aus den Fraunhofer-Instituten.

Measurement and analysis of local pulse transit time for emotion recognition

: Beckmann, Nils; Viga, Reinhard; Dogangün, Aysegül; Grabmaier, Anton


IEEE Sensors Journal 19 (2019), No.17, pp.7683-7692
ISSN: 1530-437X
ISSN: 1558-1748
Bundesministerium für Bildung und Forschung BMBF (Deutschland)
16SV7110; PAnalytics
Journal Article
Fraunhofer IMS ()
emotion; emotion recognition; pulse transit time (PTT); photoplethysmography (PPG); wearable

Emotion recognition based on physiological parameters is a research field that pushes forward from lab settings to real-life investigations. Wearable devices facilitate this advance. However, these devices are still functionally limited compared to stationary medical devices. Our goal is to extend the capability of wearable devices by developing a method that measures the pulse transit time (PTT) locally. The PTT is an interesting parameter regarding emotion recognition. A method for local PTT measurement can be implemented using two photoplethysmography (PPG) sensors. However, this method is error-prone. In this paper, the physiological background that is presumably responsible for erroneous PPG-based PTT measurements is discussed. We present an algorithm that is capable to handle the derived physiological effects. The algorithm analyzes and compares the two PPG-signals to adapt to time-varying physiological effects. By using this algorithm, calculating and analyzing of the local PTT in the context of emotion recognition become possible. A study ( n=40 ) to test the algorithm and investigate the usefulness of local PTT analysis for emotion recognition in combination with other physiological signals was conducted. PTT-based parameters, which were derived from the frequency domain of the signal, showed a statistically significant ( p<0.05 ) difference between induced emotional states, if calculated by the developed algorithm. Our findings indicate that parameters derived by our method are significantly affected by emotional stimuli. We suggest that this method can be used to advance emotion recognition investigations in real life as it can potentially be integrated into a single wearable device.