Taking it at face value: confound or defining feature?

Abstract

The aim of my thesis was to investigate the information hidden in an emotional face, specifically happy and angry faces. To do this I used data gained from a machine learning algorithm. In this thesis I defined happy and angry faces by using visual features. However, these visual features can sometimes be influenced by context dependent factors. Researching emotional faces has historically been difficult due to context factors being difficult to filter out. Therefore in this thesis the emphasis is placed on finding visual features that describe these emotional faces and if these visual features are context dependent.