Citation

Shimizu S, Ito T, Yin Y, Arakawa S, Sawada O, Aoyagi I. Trans. Soc. Automot. Eng. Jpn. 2019; 50(2): 505-510.

Copyright

(Copyright © 2019, Society of Automotive Engineers of Japan)

DOI

10.11351/jsaeronbun.50.505

PMID

unavailable

Abstract

How to estimate the driver's emotion is an important topic and has attracted much attention. Some basic research results have been reported in literature, but many problems are still unsolved especially towards the practical application. In this paper, we propose a new method to estimate the driver's emotion in positive, negative and neutral state via deep learning by driver's biological signals, where the ECG(ElectroCardioGram) data sequence is changed to an image. The estimating accuracy is up to 90% by our initial experiment results. Furthermore, we show the features of emotion with the Grad-CAM method and find that they appear in T and U wave of ECG instead of R wave, which means a new explanation.

Language: ja

Keywords

deep learning; human engineering; driver condition; physiological measurement