Continuous emotion estimation for human machine interaction

Title

Continuous emotion estimation for human machine interaction

Creator

Lee H.-S.; Xu M.; Karuppusamy N.S.; Kang B.-Y.

Description

Humans are able to interact and bond very efficiently with other species because every living organism has some form of emotion in them. Due to the advances in science and technology human life has become more dependent on machines for better living. The recent advances in technology enabled machines to become smarter but not efficient in terms of interaction with humans. Hence to address this issue and to bridge the gap between human machine interactions we propose a system to estimate human emotions from facial expressions. We believe that facial expressions are a form of nonverbal communication and primary means of conveying information. The system uses linear regression model to calculate emotional state of a facial expression which is mapped onto continuous 2-D coordinates with arousal and valence as axis from a captured digital image. Thus the proposed method estimates emotion continuously and predictively like humans rather than classifying the emotions because emotions are continuous and they have many dimensions. By estimating emotions continuously machines can better interact with humans. Experimental results showed that our system provides superior predictive performance. 2015 American Scientific Publishers. All rights reserved.

Source

Advanced Science Letters, Vol-21, No. 3, pp. 404-407.

Date

2015-01-01

Publisher

American Scientific Publishers

Subject

Arousal; Emotion recognition; Face detection; Linear regression; Valance

Coverage

Lee H.-S., School of Mechanical Engineering, Kyungpook National University, Daegu, 702-701, South Korea; Xu M., School of Mechanical Engineering, Kyungpook National University, Daegu, 702-701, South Korea; Karuppusamy N.S., Department of Mechanical Engineering, Christ University, Bangalore, 560029, Karnataka, India; Kang B.-Y., School of Mechanical Engineering, Kyungpook National University, Daegu, 702-701, South Korea

Rights

Restricted Access

Relation

ISSN: 19366612

Format

Online

Language

English

Type

Article

Identifier

https://doi.org/10.1166/asl.2015.5773

https://www.scopus.com/inward/record.uri?eid=2-s2.0-84941133918&doi=10.1166%2fasl.2015.5773&partnerID=40&md5=dc6873d6bd593944c25d735827a2abbe