Emotion Recognition of Museum Visitors during Art Appreciation Based on Facial and Behavioral Features

Abstract

In recent years, museums have been increasingly expected to facilitate communication in order to meet the intellectual needs of visitors. Although museum educators—experts in learning within museums and art galleries—have been employed to fulfill this role, there remains a persistent shortage of personnel. While systems such as audio guides are anticipated to serve as substitutes, a significant gap still exists between these systems and human guidance. Therefore, this study proposes a method to estimate the emotional state of museum visitors, aiming to provide dynamic guidance tailored to individual interests and concerns. To establish a method for emotional state estimation, we conducted a data collection experiment using a newly developed guidance system equipped with sensing capabilities at the Ohara Museum of Art, involving 30 participants. Based on the collected data of facial expressions and gestures, we constructed and evaluated an emotional state estimation model. As a result, the binary emotion classification model for positive and negative emotions achieved an accuracy of 64%. In contrast, regression models for satisfaction, comprehension, and perplexity did not yield effective results; however, some degree of correlation was observed, providing insights for future improvements.

Links

DOI:

code:references.bib

@inproceedings{bib:torigaki_EmoMuseum_BTIW2026,

author={Torigaki, Yohei and Matsuda, Yuki},

title={Emotion Recognition of Museum Visitors during Art Appreciation Based on Facial and Behavioral Features},

booktitle={The 3rd Academy of Behavior Transformation by AIoT International Workshop (BTIW '26)},

pages={x--x},

year={2026},

doi={},

url={}

}

https://scrapbox.io/files/698df61e453671c306223bcf.png

Category

Conference

Project

Keywords