Abstract
This paper discusses improvement of the method for detecting biomechanical distraction via analysis of static pressure distribution on the driving seat. The previous research showed that a detection method, which extracts the Higher-order Local Auto-Correlation (HLAC) feature from a single pressure distribution, was not high enough as a reliable method. Based on detailed review of the result, we propose combining sensors on the seat cushion and the backrest instead of using just the sensor on the seat cushion, and making the learning data set which has wide varieties for robust recognition. The result of an experiment shows that the mean recognition rate was approximately 85% suggesting the effectiveness of the detection method.
