Emotion pattern detection on facial videos using functional statistics

TL;DR

This paper introduces a functional statistical approach using Functional ANOVA and F-tests to analyze facial muscle movement patterns over time, aiming to improve automatic emotion recognition from videos.

Contribution

It presents a novel application of Functional ANOVA and F-tests for detecting emotion-related facial expression patterns in videos, advancing automatic emotion recognition methods.

Findings

Identified significant facial muscle movement patterns associated with different emotions.

Demonstrated the effectiveness of functional F-tests in distinguishing emotional expressions.

Laid groundwork for developing reliable automatic emotion recognition systems.

Abstract

There is an increasing scientific interest in automatically analysing and understanding human behavior, with particular reference to the evolution of facial expressions and the recognition of the corresponding emotions. In this paper we propose a technique based on Functional ANOVA to extract significant patterns of face muscles movements, in order to identify the emotions expressed by actors in recorded videos. We determine if there are time-related differences on expressions among emotional groups by using a functional F-test. Such results are the first step towards the construction of a reliable automatic emotion recognition system