Real-time detection of driver distraction: random projections for pseudo-inversion-based neural training

摘要: There is an accumulating evidence that distracted driving a leading cause of vehicle crashes and accidents. In order to support safe driving, numerous methods detecting distraction have been proposed, which are empirically focused on certain contexts gaze behaviour. This paper aims at illustrating method for the non-intrusive real-time detection visual based dynamics data environmental data, without using eye-tracker information. Experiments carried out in context automotive domain European project Holides, addresses development qualification adaptive cooperative human–machine systems, co-funded by ARTEMIS Joint Undertaking Italian University, Educational Research Department. The collected analysed single-layer feedforward neural network trained through pseudo-inversion methods, characterized direct determination output weights given randomly set input biases. One main feature our work convenient setting so-called sparse random projections: presence great number null elements involved matrices makes especially parsimonious use run time network. Moreover, we genetic approach better explore space. obtained results show performance with respect classical effective memory resources.