Analysis of Driver Stress Coping Styles Caused by Driving Environment and Driving Characteristics

Tomoro Okajima･Kent Nagumo･Akio Nozawa (Aoyama Gakuin University)

In this study, we focused on two types of stress coping responses that people exhibit to external stress stimuli: active and passive coping responses. These responses are considered to be effective as indicators for evaluating driver safety and comfort. We analyzed stress coping responses of drivers caused by driving environment and driving characteristics based on hemodynamics and suggested its effectiveness as indicators for evaluating driver safety and comfort.

Estimating the Gazing Point of Drivers using Machine Learning with High Generalizing Accuracy

Yui Miyoshi･Yuji Matsuki (Fukuoka Institute of Technology)

In our previous study, the authors developed a method to estimate a driver's gaze point on a driving simulator screen using machine learning. This method utilized facial images captured by three cameras. However, its performance was evaluated with only one participant, and the study did not consider its generalization accuracy. In this study, we have enhanced the model to achieve high generalization accuracy using data captured from 11 experimental participants.

Measurement Method and Evaluation of Occupant's Back during Seating

Hotaka Wakasugi･Shuta Imai･Nobuaki Nakazawa･Shinya Okamoto･Hisato Fukuda (Gunma University)･Tsutomu Iwase (Gunma University/SUBARU)･Shunpei Nakamura･Kyohei Uchikata･Masami Handa･Yusuke Takagi (SUBARU)

This study proposes a method for measuring occupant posture during seating and its application. We developed a measurement system that combines the measurement of spinal column shape using an accelerometer and the estimation of the relative position of the spinal column to the seat surface using image processing. Using this system, various seating postures on a car seat were measured and evaluated.

Generality Evaluation of Human Behavior Model in Driving Based on Recurrent Neural Network

Suzuka Seki･Jun Ishikawa (Tokyo Denki University)

This paper reports evaluation results of the generality of a human behavior model for driving based on a recurrent neural network (RNN) that the authors have proposed. Specifically, the evaluation was conducted by simulation to check whether the RNNs trained with M sequences, the amplitude of which is ±1, can reproduce appropriate responses for different amplitudes and smooth trajectories. As a result, the generality of the RNN models was confirmed.