On Shape and the Computability of Emotions

Area of Science:

• Psychology
• Computer Science
• Visual Arts

Background:

• Shape characteristics like roundness, angularity, simplicity, and complexity are theorized to influence human emotional responses.
• Previous research has not modeled the dimensional aspects of emotions evoked by shape features, particularly roundness and angularity.

Purpose of the Study:

• To statistically analyze the relationship between shape features in natural images and the emotions they elicit in humans.
• To model emotions dimensionally, predicting valence and arousal ratings for enhanced emotional content analysis.
• To differentiate emotionally charged images from neutral ones with high accuracy.

Main Methods:

• Utilized the International Affective Picture System (IAPS) dataset for experimental analysis.
• Developed and applied statistical models to quantify the influence of shape features (roundness-angularity, simplicity-complexity) on emotional content.
• Integrated shape features with existing state-of-the-art image features to improve prediction and classification performance.

Main Results:

• Provided statistical evidence for the significant impact of roundness-angularity and simplicity-complexity on predicting image emotional content.
• Demonstrated improved prediction and classification accuracy by combining shape features with other advanced features.
• Successfully modeled dimensional emotional responses (valence and arousal) rather than discrete categories.

Conclusions:

• Shape features in natural images are significant predictors of emotional content.
• Dimensional modeling of emotions based on visual features offers advantages over discrete emotional categories.
• The study accurately distinguishes emotionally evocative images from neutral ones, with implications for affective computing and visual media analysis.