Estimation of driver's posture using pressure distribution sensors in driving simulator and on-road experiment

Driving assistant is a very effective technology to reduce the traffic accidents by warning driver and supporting the driving operation. In order to give more accurate assistant, both outside condition and inside driver's situation have to be measured correctly and speedily. In this research, we propose a machine Learning based method to estimate the driving posture only from the pressure distribution between driver and car driving seat using a type of thin and soft sensor sheet. Using this method, drivers posture can be obtained without wearing any sensor on drivers' body or setting up any camera or motion capture device in the narrow car inside space. The pressure distributions of different driving postures were measured both in a driving simulator and on-road experiment. Using the measured data, we create the classifiers based the supervised machine learning method — Support Vector Machine (SVM) and tested the generated classifiers for continued driving motions. The effectiveness of this method has been confirmed by comparing the estimated driving postures with the manually labeled postures.