Using a humanoid robot for assessing vocal emotion recognition

Authors: Gloria Araiza Illan1, Luke Meyer1, Laura Rachman1, Etienne Gaudrai2, Deniz Başkent1

1 University Medical Center Groningen
2Université de Lyon

Background: Vocal emotion recognition relies on multiple factors, from the perception of vocal cues that express emotions, to their correct categorisation and labelling. We have previously shown that a humanoid NAO robot is a feasible interface for different voice and speech perception tests; as a next step, we propose the robot as an alternative interactive and engaging testing interface for vocal emotion recognition.

Methods: We present a study with 28 young normal-hearing adults using the EmoHI vocal emotion recognition test implemented in two interfaces: a computer and a NAO robot to study the feasibility of the latter to conduct such a test. The EmoHI test, developed by our group as a simple task with three emotions (angry, happy, sad), is to be used with children and adults of different hearing status. Participants performed the EmoHI test twice, once with the computer and once with the NAO. Test results and durations were compared between interfaces. Engagement was assessed via questionnaires filled in after each test. Test stimuli were presented randomly. After each stimulus, participants indicated the emotion of the voice, either by selecting the corresponding facial expression on the computer interface or by using the tactile sensors on the robot.

Results: Test results in sensitivity index, d’, with the NAO (2.02 ± 0.26) were similar to that with the computer (1.97 ± 0.39). Test durations were also similar between the NAO and computer (3.63 minutes ± 15 seconds, and 3.76 minutes ± 12 seconds, respectively). Results from the questionnaires indicated a preference for the robot, selected as a more enjoyable and engaging interface.

Conclusion: Results demonstrate the potential of the NAO robot as an interactive and engaging testing interface by showing 1) its similar functionality to the computer interface, and 2) participants’ preference for the robot over the computer.