Analysis of hands activity for automatic driving risk detection

In this paper a novel methodology to measure driving risk based on hands activities and other driving variables is presented. The proposed methodology has been developed and tested in several driving sessions executed in two highly realistic simulators by several professional and non-professional drivers. The driver’s hand positions are used to feed a risk buffer that is in charge of penalizing wrong hands activities and praising good hands activities to generate a measure of the driving risk. In addition, other relevant variables are considered to increase the risk level (activators) or decrease the risk level (inhibitors) depending on the driving situations. In order to estimate the parameters of the risk buffer, a genetic algorithm (GA) and a driving risk ground truth acquired from traffic safety experts were considered. Several experiments have been designed and executed in truck and vehicle simulators. The proposed methodology has been tested in different driving scenarios. On the one hand, our method provides an estimation of the most relevant variables used for the traffic safety experts to define the driving risk. On the other hand, our method shows its effectiveness to automatically detect risky situations related to bad driver’s behavior. In the future, the present system will be integrated in a global alarm system that will be included in real vehicles for the improvement of traffic safety.

► We propose a methodology to measure driving risk based on driver’s hand activities.
► A risk buffer is built by penalizing/praising wrong/right hand activities.
► The risk model is generated considering variables from the driver, vehicle, and road.
► We developed and tested our system on two realistic simulators: truck and vehicle.
► Our method detects risky situations related to bad driver’s hand behaviors.