Facial Components and Features Points Extraction
For Human Emotions Recognition

The methods presented in this document aim at extracting keypoints from human faces in order to recognise emotions. The feature points are part of the MPEG-4 standard. Faces are supposed to be detected from background prior to the present process. We use a hierarchical approach. Main components (high level) of the face are first exctracted (eyebrows, eyes, nose and mouth). Keypoints (low level) are then localised from the detected features.
Figure1: Features points extraction

Original image Features points

Features extraction and Feature Points Localisation:

First approach: Using morphological operators
Second approach: Using faces symmetrical properties


Some results there


PHYSTA Project
Face Recognition Homepage