On the base of information about body position, movement, shape and change and its parts we can derive facts about a driver´s activity, his mental state and competence for driving a car. Information about the driver´s possible health obstacles can contribute to the higher safety.
It is obvious that states like momentary drowsiness, epileptic fit or heart attack shift the driver´s attention from driving, this could be very dangerous. In such a case it would be appropriate to recognize these states and swich on assistant systems in a car that are able to hand over the driving instead of the driver and prevent thus from potential dangerous traffic situations. Currently we solve the issue thanks to deep neuron nets based on architecture encoder-decoder (the research is still going in the partnership with ŠKODA AUTO a.s.)