Optimization of categorizing driver´s head motion for driving assistance systems

Many car accidents are caused by driver´s deviation from normal condition like carelessness. We aim to construct a driving assist system that is able to detect driver´s deviation signal from normal condition. The system detects the deviation signal using driver´s time-series head motion information. In this paper, we optimize categorization of driver´s head motion using two kinds of unsupervised neural networks: Self-Organizing Maps and Fuzzy Adaptive Resonance Theory, and discuss the relation between vigilance parameter and integrated categories.