Abstract
Over the last decade, an increasing number of studies have focused on automated recognition of human emotions by machines. However, performances of machine emotion recognition studies are difficult to interpret because benchmarks have not been established. To provide such a benchmark, we compared machine with human emotion recognition. We gathered facial expressions, speech, and physiological signals from 17 individuals expressing 5 different emotional states. Support vector machines achieved an 82% recognition accuracy based on physiological and facial features. In experiments with 75 humans on the same data, a maximum recognition accuracy of 62.8% was obtained. As machines outperformed humans, automated emotion recognition might be ready to be tested in more practical applications.
Notes
Background. This article is partly based on the Ph.D. thesis of Joris H. Janssen (Eindhoven University of Technology) and the M.Sc. thesis of Paul Tacken (Radboud University Nijmegen).
Acknowledgments. We thank internal reviewers and three anonymous reviewers for their valuable remarks, detailed comments, and constructive feedback.
Support. This publication was supported by the Dutch national program COMMIT (projects P4 Virtual worlds for well-being and P7 SWELL).
HCI Editorial Record. First manuscript received September 6, 2011. Revision received August 16, 2012. Final manuscript received October 16, 2012. Accepted by Andrew Monk. — Editor