A multi-modal approach for emotion recognition of TV drama characters using image and text

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

11 Scopus citations

Abstract

Research on facial emotion recognition has long been popular for various purposes. This paper investigates the recognition of the character emotions, to assist in understanding the story. The goal of this research is to classify the facial images of the characters in the Korean TV series 'Misaeng: The Incomplete'1 into 7 emotions: Angry, Disgust, Fear, Happy, Neutral, Sad, and Surprise. We built a multi-modal deep learning model which utilizes facial images as well as textual information describing the situations, to classify the facial images. Our experiments indicate that employing multi-modality enhances the performance of facial emotion recognition of story characters.We concludes with discussions and future work.

Original languageEnglish
Title of host publicationProceedings - 2020 IEEE International Conference on Big Data and Smart Computing, BigComp 2020
EditorsWookey Lee, Luonan Chen, Yang-Sae Moon, Julien Bourgeois, Mehdi Bennis, Yu-Feng Li, Young-Guk Ha, Hyuk-Yoon Kwon, Alfredo Cuzzocrea
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages420-424
Number of pages5
ISBN (Electronic)9781728160344
DOIs
StatePublished - Feb 2020
Event2020 IEEE International Conference on Big Data and Smart Computing, BigComp 2020 - Busan, Korea, Republic of
Duration: 19 Feb 202022 Feb 2020

Publication series

NameProceedings - 2020 IEEE International Conference on Big Data and Smart Computing, BigComp 2020

Conference

Conference2020 IEEE International Conference on Big Data and Smart Computing, BigComp 2020
Country/TerritoryKorea, Republic of
CityBusan
Period19/02/2022/02/20

Keywords

  • Asian-face
  • CNN
  • Contextualized-embedding
  • Face-emotion-classification
  • Multi-modal

Fingerprint

Dive into the research topics of 'A multi-modal approach for emotion recognition of TV drama characters using image and text'. Together they form a unique fingerprint.

Cite this