Multimodal human interaction analysis in vehicle cockpit
Résumé
Nowadays, every car maker is thinking about the future of mobility. Electric vehicles, autonomous vehicles and sharing vehicles are one of the most promising opportunities. The lack of authority in autonomous and sharing vehicles raises different issues from which one of the main issues is passenger safety. To ensure it, new systems able to understand interactions and possible conflicts between passengers have to be designed. They should be able to predict critical situations in the car cockpit, and alert remote controllers to act accordingly. In order to better understand the features of these insecure situations, we recorded an audio-video dataset in real vehicle context. Twenty-two participants playing three different scenarios (“curious”, “argued refusal” and “not argued refusal”) of interactions between a driver and a passenger were recorded. We propose a deep learning model to identify conflict situations in a car cockpit. Our approach achieves a balanced accuracy of 81%. Practically, we highlight the importance that combining multimodality namely video, audio and text as well as temporality are the keys to perform such accurate predictions in scenario recognitinn.