Emotional Expression Classification using Time-Series Kernels

TL;DR

This paper demonstrates that kernel methods applied to 3D facial landmark motion data can classify emotional expressions with high accuracy, even from very short video segments, advancing facial expression analysis techniques.

Contribution

It introduces a kernel-based approach using dynamic time warping for classifying facial expressions from 3D landmark motion data, achieving high accuracy with minimal frames.

Findings

Over 99% accuracy in classifying facial expressions.

Recognition of expressions within 200 milliseconds from onset.

Effective use of PCA-compressed shape parameters for classification.

Abstract

Estimation of facial expressions, as spatio-temporal processes, can take advantage of kernel methods if one considers facial landmark positions and their motion in 3D space. We applied support vector classification with kernels derived from dynamic time-warping similarity measures. We achieved over 99% accuracy - measured by area under ROC curve - using only the 'motion pattern' of the PCA compressed representation of the marker point vector, the so-called shape parameters. Beyond the classification of full motion patterns, several expressions were recognized with over 90% accuracy in as few as 5-6 frames from their onset, about 200 milliseconds.