Authors:
Braj Bhushan
1
and
Prabhat Munshi
2
Affiliations:
1
Department of Humanities & Social Sciences, Indian Institute of Technology Kanpur, India
;
2
Nuclear Engineering and Technology, Indian Institute of Technology Kanpur, India
Keyword(s):
Facial Expressions, Geometric Feature Extraction, Fractal Dimension, Root Mean Square Pixel Difference, Self-similarity.
Abstract:
Fractal dimension analysis of the images of facial expressions has been reported earlier by Takehara and colleagues We have performed a similar exercise for two Indian databases, the Indian dataset of basic emotions and the Indian Affective Picture Database, to examine the relationship between the geometric properties of the facial expressions vis-à-vis the intensity of expressions and the viewing angle. It is a first of its kind in the Indian context. We analyzed the geometric pattern of three regions of the face, computed pixel difference, and calculated fractal dimensions of the expressions for all the images of these two databases. Thereafter, we analyzed the obtained outcomes of the geometric analyses and the reported unbiased hit rates for these databases, respectively. Results suggest that recognition of facial expressions is independent of the viewing angle. Further, happiness and anger are recognized best irrespective of their intensity followed by more intense surprise and
disgust. The Root Mean Square pixel difference shows identical pattern in the expressions of happiness and disgust. Fractal dimensions indicate self-similarity among surprise, happiness, and disgust.
(More)