Fraunhofer-Gesellschaft

Publica

Hier finden Sie wissenschaftliche Publikationen aus den Fraunhofer-Instituten.

Emotion recognition in user-centered design for automotive interior and automated driving

 
: Bischoff, Sven; Ulrich, Christian; Dangelmaier, Manfred; Widlroither, Harald; Diederichs, Frederik

Binz, Hansgeorg (Hrsg.); Bertsche, Bernd (Hrsg.); Bauer, Wilhelm (Hrsg.); Spath, Dieter (Hrsg.); Roth, Daniel (Hrsg.) ; Univ. Stuttgart, Institut für Konstruktionstechnik und Technisches Design; Univ. Stuttgart, Institut für Maschinenelemente; Univ. Stuttgart, Institut für Arbeitswissenschaft und Technologiemanagement -IAT-; Fraunhofer-Institut für Arbeitswirtschaft und Organisation -IAO-, Stuttgart:
Stuttgarter Symposium für Produktentwicklung, SSP 2017 : Produktentwicklung im disruptiven Umfeld; Stuttgart, 29. Juni 2017, Wissenschaftliche Konferenz
Stuttgart: IRB Mediendienstleistungen, 2017
10 S.
Stuttgarter Symposium für Produktentwicklung (SSP) <2017, Stuttgart>
Englisch
Konferenzbeitrag
Fraunhofer IAO ()

Abstract
In this paper an experiment with a sensor platform for measurement of emotions in the automotive context will be presented. Based on state-of-the-art equipment for emotion recognitions, an architecture for an emotion recognition sensor platform was specified and tested for specific use cases and target emotions in the context of automotive Human-Machine-Interaction (HMI) the car interior and automated driving maneuvers. The sensors include facial expression analysis, head and eye tracking, GSR, EEG and heart rate measurement. Emotional target states are joy, pleasantness, anger, sadness, fear, cognitive workload, visual distraction, boredom, comfort (ergonomic), trust (in automation), interestedness, (dis-)satisfaction (with product features), and surprise. Within a driving simulator study with 30 participants those emotions could be evoked with the designed stimuli at 90% success rate. Many classification results of the different sensors correlate. Results revealed no absolute redundancy in different sensors. With classification rates of the 28 %, Random Forest classification achieved the best results in predicting target emotional states (5% random guess).

: http://publica.fraunhofer.de/dokumente/N-461359.html