DescripciĆ³ del projecte
The goal is to design a complete system to evaluate the drivers’ capabilities for secure driving. To objectively predict the ability to drive the system should collect multimodal data from the driver including cognitive state detected by psychophysical tests, motor skills collected by wearable sensors as well as visual attention and driving behavioural markers from video analysis.
In order to Extract Biometrical and Behavioural Data, we will apply video analysis and deep learning methods developed to classification of the driver behavioural patterns such as facial expression recorded in videos and biometrics to extract indicators of the driver current state. In particular, we will combine indicators assessing patient biometrics and behavioural data in a multi-view data model. To account for uncertainty and personalization, deep networks will be trained using a multi-objective approach to achieve a compromise between classification accuracy and uncertainty in predictions. A personalized driving capability index will be built out of the combination of the resulting computations. We plan to analyse its dynamic behaviour in front of different risky situations and provide a final global evaluation.
The developed techniques will be applied to evaluate the capability to drive for patients with ParkinsonĀs Disease , as well as, incorporated into Continental security systems for drivers’ monitoring.
The project will be carried in cooperation with the following two institutions:
-Computer Vision Center in the Universitat AutĆ²noma de Barcelona.
-Continental automotive Romania SRL
