Presented at First international conference on driver distraction and inattention (DDI 2009), Gothenburg, Sweden, September 28-29, 2009

This paper proposes a method to identify driver posture based on pressure distribution on the driving seat. In our method, the Higher-order Local Auto-Correlation (HLAC) features are extracted from an image of a pressure distribution. We conducted an experiment to investigate the effectiveness of our method. The data were collected on a driving simulator. The results show that the method is potentially useful for estimating driver actions. We also tried to find ways to improve the performance of the method. The results show that using two sensor sheets on the seat cushion and the backrest is necessary. The resolution of a sensor sheet can be reduced to half of the original or less. If the training samples have lots of variations, the mean recognition rate goes up to approximately 85%, suggesting the effectiveness of the detection method.