Recognising Facial expression in video

📄 Project Abstract

This project set out to develop a method of recognising facial expression in video footage. The video format used was a webcam scenario, whereby the camera was close to the face, and occupied most of the image space. The method being tested was Principal Component Analysis. Movements of salient facial features were tracked, and PCA performed on the movements of these points in footage featuring various facial expressions. Videos were made of sufficient subjects to encapsulate a range of face and expression types. It was hoped that PCA would yield variations in components consistent across subjects and expressions, allowing prototype values for expressions to be established. Frames could then be assigned expressions based on similarity to prototype values obtained. This was to form the basis of the recognition system. However, it was found that different subjects produced different PCA values for expressions, and that there were no consistent features which would allow a recognition sys...