Synthetic Emotions for Humanoids: Perceptual Effects of Size and Number of Robot Platforms

The recognition and display of synthetic emotions in humanoid robots is a critical attribute for facilitating natural human-robot interaction. The authors utilize an efficient algorithm to estimate the mood in acoustic music, and then use the results of that algorithm to drive movement generation systems to provide motions for the robot that are suitable for the music. This system is evaluated on multiple sets of humanoid robots to determine if the choice of robot platform or number of robots influences the perceived emotional content of the motions. Their tests verify that the authors’ system can accurately identify the emotional content of acoustic music and produce motions that convey a similar emotion to that in the audio. They also determine the perceptual effects of using different sized or different numbers of robots in the motion performances. DOI: 10.4018/jse.2012070104 International Journal of Synthetic Emotions, 3(2), 68-83, July-December 2012 69 Copyright © 2012, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited. they must be able to extract certain high-level features from the audio and incorporate those features into their performances. It is not sufficient for the robots to simply perform a series of pre-recorded or choreographed motions. Live performances inevitably vary from one show to the next, and even if there are no major changes in the music, there will always be small shifts in tempo, phrasing, and other factors. If the robot can only play according to a predetermined sequence, its responses will not necessarily mesh with the live performance. Additionally, we do not want the musicians to be forced to play a certain way in order to conform to the robot. The robot, rather, should be capable of following the musicians, even if their performance is completely different from previous ones. We have performed substantial prior work on enabling robots to step or dance in response to music (Grunberg, Lofaro, Oh, & Kim, 2011; Kim et al., 2010). These performances, however, only considered the beat locations and tempo of the audio, and disregarded other aspects. We now turn our attention to another crucial feature: the emotional content, which we also refer to in this paper as ‘mood’, of music. Humans modify their dance motions so that their performances match the mood of the music they dance with, and when the dance motion and musical audio do not convey the same emotion, the resulting performance can look jumbled or confused. Therefore, to optimize their dances, the robots should be capable of determining an appropriate mood and then communicating it via gestures. The choice of robot platform will influence the types of motions and emotions that the robot is able to produce. Smaller humanoids with more cartoonish looks for example, may be more suitable for happier or ‘cute’ emotions. They are also generally more rugged and easier and cheaper to repair than larger humanoids, so they offer practical advantages as well. Conversely, adult sized humanoids are often more human-like in appearance, so they may be able to display gestures with a higher degree of fidelity than the cartoonish miniature platforms. These robots also tend to boast more computational power and more sophisticated motion algorithms. As both types of robots offer different advantages, we would like to determine which ones are better at communicating which emotions to human audiences. Another important decision is the number of robots to be used in a performance. Some human dance troupes have several members perform identical or similar motions to communicate the desired emotional effect to their audience. Using multiple robots instead of one might increase some of the emotional intensity, but it could also encourage the viewer to focus less specifically on any one single robot and thus be less moved by whatever emotion is being expressed. We would therefore like to more precisely identify if using multiple humanoids for dance performances influences how humans perceive the mood of the dances.