Face Reconigtion, Quantification and Measurement of Facial Expressions with Facial Landmarks

Human beings have always used two main modes to communicate verbal and non-verbal ways that use words and facial expressions respectively. The human face provides a rich source of information and is associated with the signalling of emotions and pain (Ekman, 1993), the personality traits, the communicating of emphatic understanding (Bavelas et al., 1986) and the regulation of conversations (Cohn and Elmore, 1988). Furthermore, emotions are fundamental to human lives and decision-making. Therefore, variation of the human face has been a subject of interest since ancient times, and in recent years, we have witnessed the rapid emergence of an interest for the automated analysis and interpretation of facial activity through computer vision. Long before modern scientific research, artists often accurately depicted variations in human physiognomy. The earliest recorded set of facial proportional tenets was formulated in Greek classical canons around 450 B.C. These tenets were further elaborated by European Renaissance artists, particularly Leonardo da Vinci, into a system known as the “neoclassical canons” (Farkas et al.). Thus, the face has always been the centre of attention during our collective and social interaction due to its communicative power. In fact, facial expressions and the changes in facial aspects allow us to know the psychological and emotional conditions of people around us and help to define social interactions and conversations. For this reason, in the last few years, face perception and face processing have become significant subjects of research by psychologists, sociologists, scientists and by researchers in computer graphics. In recent years, we have witnessed the rapid emergence of an interest for the automated analysis and interpretation of facial activity through computer vision. Since the beginning of the 1990s, the subject has become a major issue, mainly due to the important real-world applications of face recognition like smart surveillance, secure access, telecommunications, digital libraries and medicine. Humans has always dreamed of building machines that are as human as possible, so facial recognition is a basic central point for building a human-like machine. Furthermore, recognizing emotions is the first step to making a computer affectively intelligent. The automation of human face processing by a computer will be a very important step on the way to bring a radical change and develop an efficient human-machine interface. Face processing by machines could completely change areas as medicine, law, education, marketing, etc. and is useful, as mentioned above, for various fields of application, including behavioral research, affective computing, 3D facial reconstruction and animation, car safety, social security, health care and others. This research presents a methodological contribution to the study of facial expression of emotion. It describes a technique developed to quantify facial expressions and proceeds to test its validity as a measurement tool. This thesis describes the use of a measurement technique for the study of the facial expression of emotion and illustrates a system for observing and measuring the "action units" of a face using photos as input. The aim is to quantify facial movements and measure distance scores between facial landmarks made by a set of basic categories of emotions and neutral expressions (between the current face image showing an emotion and the referenced neutral face) consistently.