We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

cs.CV

Change to browse by:

References & Citations

DBLP - CS Bibliography

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Computer Science > Computer Vision and Pattern Recognition

Title: Context-Dependent Models for Predicting and Characterizing Facial Expressiveness

Abstract: In recent years, extensive research has emerged in affective computing on topics like automatic emotion recognition and determining the signals that characterize individual emotions. Much less studied, however, is expressiveness, or the extent to which someone shows any feeling or emotion. Expressiveness is related to personality and mental health and plays a crucial role in social interaction. As such, the ability to automatically detect or predict expressiveness can facilitate significant advancements in areas ranging from psychiatric care to artificial social intelligence. Motivated by these potential applications, we present an extension of the BP4D+ dataset with human ratings of expressiveness and develop methods for (1) automatically predicting expressiveness from visual data and (2) defining relationships between interpretable visual signals and expressiveness. In addition, we study the emotional context in which expressiveness occurs and hypothesize that different sets of signals are indicative of expressiveness in different contexts (e.g., in response to surprise or in response to pain). Analysis of our statistical models confirms our hypothesis. Consequently, by looking at expressiveness separately in distinct emotional contexts, our predictive models show significant improvements over baselines and achieve comparable results to human performance in terms of correlation with the ground truth.
Subjects: Computer Vision and Pattern Recognition (cs.CV); Applications (stat.AP)
Cite as: arXiv:1912.04523 [cs.CV]
  (or arXiv:1912.04523v1 [cs.CV] for this version)

Submission history

From: Victoria Lin [view email]
[v1] Tue, 10 Dec 2019 06:10:25 GMT (1781kb,D)

Link back to: arXiv, form interface, contact.