Information Technology Reference
In-Depth Information
Japanese rated the perceived emotions of HN (happy eyes and neutral mouth)
significantly happier than Hungarians and SN (sad eyes and neutral mouth) as
significantly sadder than Hungarians. Hungarians rated the perceived emotions of NS
(neutral eyes and sad mouth) significantly sadder than Japanese. This means the
expressions where either eyes or mouth is/are neutral confirmed the hypothesis,
except NH (neutral eyes and happy mouth). However, the contradictory expressions
(HS, SH), where the eye expressions and mouth expressions show opposite
expressions, fail to show significant cultural differences.
The effectiveness of the mouth region is shown again in NH and SH (both with
happy mouths). Both countries' perceived emotions of NH have the happiest scores.
SH's score are higher than HN.
Third, we focused on the differences in ratings by the shape of the eye/mouth when
the other facial region was fixed. Both countries responded to the differences in the
mouth shape as significantly and dynamically as expected. However, Japanese
responded to the differences of the eye shape more significantly and dynamically than
Hungarians, implying that Japanese focused more strongly on the eyes even when the
mouth region has the most convincing happy expression; Hungarians continued to
focus on the mouth region.
However, the differences in the eye shape with a sad mouth did not result in the
expected direction. Perhaps the reason reflects the nature of the effectiveness of the
mouth region in facial expression recognition and eye designs. Some subjects
commented that the happy eye designs resembled surprised expressions when isolated
from the mouth region. Although the perceived emotions of the happy eyes using
static images of the eye region in the pre-evaluation test had 90% in Japan and 85%
accuracy in Hungary, the result implies we must design the eyes more carefully when
they are used solely apart from other facial regions.
Former studies
[10, 11]
used emoticons and photorealistic human facial
expressions. Emoticons can convey emotions simply and effectively because we use
them in our daily lives. Since the photorealistic human facial expressions used in
these studies were made by professional actors, they naturally and effectively
conveyed the intended emotions. However, cartoonish facial designs have wide
variations caused by the designers' drawing styles and techniques. Although we used
CharToon as a design tool to minimize possible variations, the agent faces in the
experiment had more room for improvement.
Another consideration should be put on cultural diversity in designing cartoon
faces. Koda reports in [9]
that the recognition accuracy of facial expressions is higher
for virtual agents designed by the same cultural group as the subjects. The facial
expressions used in the experiment were designed by Japanese, thus the drawings
followed Japanese stylistic conventions for expressing emotions, and Japanese have
more exposure and experience in such drawings. Although the pre-evaluation of facial
expressions did not show significant cultural differences, we cannot exclude the
possibility that cultural differences in recognizing expressions in cartoon faces might
be a result of cultural differences in drawing cartoon faces. We had an interesting
result from the pre-evaluation of the agent faces. Hungarians' perceived age of the
agent faces used in the experiment are much younger than the Japanese ones.
Hungarian perceived the age of the agent faces as in their teens or younger, and
Japanese as in their twenties.
