Internal state estimation by thermal image and identification of face and nose position

There is a close relationship between body tem-perature and activity, such as decreased body temperature when drowsiness occurs in humans. Therefore, we can estimate drowsiness by measuring changes in human body temperature. In particular, this paper focuses on the facial area that includes the fore...

Full description

Saved in:
Bibliographic Details
Published in2021 Asia-Pacific Signal and Information Processing Association Annual Summit and Conference (APSIPA ASC) pp. 386 - 391
Main Authors Watanabe, Yuta, Manabe, Yoshitsugu, Yata, Noriko
Format Conference Proceeding
LanguageEnglish
Published APSIPA 14.12.2021
Subjects
Forehead
Nose
Particle measurements
Position measurement
Skin
Temperature measurement
Temperature sensors
Online AccessGet full text

Cover

More Information
Summary:There is a close relationship between body tem-perature and activity, such as decreased body temperature when drowsiness occurs in humans. Therefore, we can estimate drowsiness by measuring changes in human body temperature. In particular, this paper focuses on the facial area that includes the forehead, which is thought to reflect the temperature inside the body, and the area around the nose, where changes in skin temperature are significant. At that time, the body temperature is measured non-contact using an infrared camera without attaching electrodes or sensors. The measurement is performed without giving stress to the user. In addition, when expanding the target to multiple people, it is necessary to detect the face positions of multiple people. However, most systems that can measure temperature without contact currently use visible images simultaneously to detect the position of the face. Therefore, in order to estimate drowsiness for multiple people, the positions of the faces and noses of multiple people should be estimated using only infrared cameras. In the experiment results, the drowsiness was estimated to be almost correct. It was shown that position estimation is effective even when it is expanded for multiple people.
ISSN:2640-0103