Shape and the Computability of Emotions

What was the research project?

This research project was a study of the characteristics of shape and their effect on human emotions. It indicates that the roundness and complexity of shapes are fundamental to understanding the emotions they provoke. The hypothesis was that curved contours lead to positive feelings and that sharp contours lead to negative feelings. The project also explored complexity of shape, which the perception of simplicity is subjective to individual experience as well as parsimony and orderliness.

How data was used

Parsimony is the minimalist structures that are used to represent the object, where orderliness is the most simplistic way of organizing these structures.

To objectively compare observations on shape and emotion, physiologists created the standard International Affective Picture System dataset by gathering data on three affecting dimensions. These were valence, arousal and dominance. These researchers looked into factors such as colour, texture, composition and semantics to understand emotions, however did not address the effect of perceptual shapes.

The perception of shapes are influenced by the the context and surrounding shapes. Representation of shapes in complex images can be highly challenging, as individual perception and emotional reaction of and to shapes may be different. Non abstract images, being made of shapes that are not simple nor regular, are much more difficult to model a research framework around.

The project seeks to automatically distinguish images with strong emotional content from those that are emotionally neutral. The data will be visualized using a dimensional representation, plotting emotions on a scale rather than categorizing them directly.

Valence represents the positive or negative aspect of emotions, arousal describes reacting to stimuli and dominance means the controlling nature of the emotion. For example, anger would be high valance, high arousal and high dominance, whereas sadness would be low valence, mid arousal and low dominance. I do not necessarily agree with this – sadness can be rather controlling if it gets out of hand – I’ll need to read more about IAPS.

According to Bradley and Lang, categorized emotions do not provide a one to one relationship between the content and emotion of an image as participants perceive different emotions in the same image.

The four shape features used were line segments, angles, continuous lines and curves. Line segments are short straight lines generated by fitting nearby pixels. This can be used to determine simplicity-complexity. Angles are the angles found between any two intersecting line segments. the number of angles and number of different angles in an images can be effectively used to describe simplicity-complexity. A high number of acute angles determines an image to be angular. Continuous lines are intersecting line segments with the same orientations. Ones with different orientations are corners or points of inflections. Corners are lower than 90 degrees, whereas intersection points are the midpoint of two angles with opposite orientations. The degree of curving in continuous lines can be used to determine the complexity of the image. Curves are a subset of continuous lines, which describe the roundness of an image. Each curve is seen as a section as a ellipse, and these are imposed on top of the image to determine where curves lie.

Outcomes

It was found that images with very high or very low arousal and valance ratings had strong emotional content. Mean values were emotionally neutral. Similarly, very simple or complex and very round or angular images had strong emotional content. Simple images with low degrees of curving also fit into this category and tended to have high arousal values.

This project had some really useful language on how to describe shape that I think could be really useful within my project. However, the rest of the paper looked at creating tuples from the data and using them to improve image searches and other computing functions, which isn’t of relevance to my project.

Lu, X. Suryanarayan, P. Adams, R., Li, J. , Newman, M. , Wang, J. On Shape and the Computability of Emotions. In: Proceedings of the 20th ACM international conference on Multimedia, Nara, Japan, 29 Oct – 2 Nov. 2012. New York:ACM [online]. Available from: http://infolab.stanford.edu/~wangz/project/imsearch/Aesthetics/ACMMM2012/lu.pdf [Acessed 12 January 2014]

Advertisements

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s