Optical Flow Analysis of Emotional Human Facial Expressions – Forensic Application

Emotional facial expressions have been suggested as indicators of a person’s intended actions when they display a set of specific facial reactions. Our aim was to investigate Optical Flow to identify specific emotions with the long-term objective to create an algorithm that would automate the process. This would remove all allegations of bias to give security personnel an objective way of screening individuals in a crowd who may require closer scrutiny. This paper describes our preliminary studies, using film clips to elicit dynamic natural facial responses. This has identified key elements that need to be evaluated further, such as gender and age of the people under observation, as well as specific facial elements that may be of greater benefit than whole facial images for the purpose of screening crowds for potential perpertrators of anti-social behaviour.