Obstacle Judgment Model of In-vehicle Voice Interaction System Based on Eye-tracking

Many obstacles have happened in in-vehicle voice user interaction (VUI) systems, which affect the drivers' safety. By studying human mental models, researchers in design have found the most threatening drivers' safety is not the Natural Language Processing (NLP) error that many technicians are only working to improve. Therefore, enhancing the collaboration between driver and in-vehicle VUI system needs to study human-computer interaction, not just to rely on technological improvement. In this study, we propose using eye-tracking technology to evaluate the impact of VUI obstacles on drivers quantitatively. Then we establish an obstacle judgment model of in-vehicle voice interaction systems based on eye-tracking and predict four types of VUI obstacles with about 93% accuracy. At last, a human-computer collaboration method for the in-vehicle voice system based on this model is discussed.