Abstract
In this study, we estimate driver's posture from a single camera video for analyzing relationship between driver's behaviors and traffic accidents. In this paper, we aim at realizing robust posture estimation of upper human body while driving. For the posture estimation, specific body parts such as a head, hands, shoulders are detected and sequentially tracked by using Particle Filter and HOG feature based recognition method. The proposed method actually applied to videos captured by a drive recorder, and the results shows effectiveness of the system.
