An efficient approach for recognizing and tracking spontaneous facial expressions

An efficient approach for recognizing and tracking human facial expressions from 2D videos is introduced. Spontaneous facial expressions are different from posed expressions in both that muscles are moved and in the dynamics of the movement. Three important aspects in the general research framework of facial expressions detection and tracking by computer are addressed here: face detection, localizing facial feature points, and matching corresponding points. The detected feature points are analysed and categorized into action units (AUs) to recognize the facial expression in each frame. In this work, these aspects are carried out automatically considering the real-time video through a single step. The main objective of this research is to track and recognize the facial expressions for a given video, which later could be used to predict a human behavior before doing an event such as a crime, angry or for being nervous. To achieve this goal, several algorithms are employed and integrated to introduce our efficient approach.