Perception of emotional body language displayed by animated characters
Student thesis: Doctoral Thesis
Virtual Environments have demonstrated effectiveness for social task training such as medical training (Anolli, Vescovo, Agliati, Mantovani, & Zurloni, 2006). These types of Virtual Environments have used emotional animated characters. Even though emotions have a strong influence on human-human interactions (Gratch, Mao, & Marsella, 2006), typical system evaluation does not assess whether human and animated emotional displays are perceived similarly by observers. Moreover, the Uncanny Valley, which is a drop in believability as characters become more realistic, threatens the assumption that emotions displayed by an animated character and a human would be interpreted similarly. Thus, it is not known how appropriate the perception to a realistic emotional animated character is. This issue is especially important for social task training which require animated characters to be perceived as social and emotional partners so that trainees would be confronted with situations comparable to real life ones. Using an approach similar to the one proposed by Nass & Moon (2000) in their work on the Media Equation, this thesis investigates how emotional body language displayed by animated characters is interpreted. A psychological experiment was conducted to investigate if emotional body language would be an appropriate way for animated characters to display emotion. This was done by comparing the interpretation of emotional body language displayed by animated characters with that by real actors. The results showed that animated body language can be accurately interpreted. However, the videos of the actor were found to be more emotional, more believable and more natural than the animated characters, whilst displaying the same emotional body language. Moreover, there was a significant difference in the number of correctly interpreted negative emotions displayed. Although, there was not a difference for positive emotions. This could be due to the physical appearance of the animated character or to the loss of micro-gestures inherent to Motion Capture technology. Thus, a second comparative study was conducted to investigate the potential causes for this drop in believability and recognition. It investigated the effect of changing the level of physical realism of the animation as well as deteriorating the quality of the emotional body language itself. Whilst no effect was found regarding the deterioration of the emotional body language, the results show that the videos of the Actor were found to be more emotional, more believable and more natural than the two animated characters. These findings have strong implications for the use of Virtual Environments for social task training.
3.35 MB, PDF document