Article Preview
Top1. Introduction
Facial Expression is usually referred to as the transformation of the human face caused by an automatic response to an emotional state which may be as a result of voluntary action, in most cases spontaneous and uncontrollable. Thus, a reflex action that requires applications of a machine or intelligent system to recognize mainly the seven prototypical expressions put up by the face under any circumstance. FE has been an active research hotspot in the fields of Computer Vision (Aneja, Colburn, Faigin, Shapiro, & Mones, 2016), Pattern Recognition (Manivannan et al., 2016; Samarasinghe, 2016), Image Classification, and related fields in recent times (Kamarol, Jaward, Kälviäinen, Parkkinen, & Parthiban, 2017; Mary & Jayakumar, 2016; Valstar et al., 2017). These studies are rapidly increasing on the account of a visible manifestation of one’s affective state, cognitive activity, intention, personality, and psychopathology as opined by (Martinez, Valstar, Jiang, & Pantic, 2017). FEs are composed of macro‐ expressions, expressed in the form of anger, disgust, fear, happiness, sadness or surprise, and other involuntary rapid facial patterns, initiating a non-rigid human face signal. Facial dynamics are computed by either observing the face signals, for example (a surprise gesture can be expressed by opening the eye wide and tightening the eyelids shows gladness) or the strain magnitude estimated using the central difference technique over the dense optical flow field detected in several regions (forehead, mouth, chin, cheek) on each subject’s face (Matthew Shreve, 2011.).
Lopes, A.T., et al opined that graphical images speak a thousand words, but a typical human face will speak louder since the FE is a window to human personality; emotions, and thoughts, thus making it play a crucial role in social communication (Lopes, de Aguiar, De Souza, & Oliveira-Santos, 2017). Note, one of the fundamental things humans learn first is, recognition of faces and interpretation of a few basic emotions from them. It is hard to imagine the expression of humor, love, appreciation, grief, enjoyment, or regret without FE. Facial recognition is one of the fastest-growing technologies, yet the least interphase with biometrics compared to other techniques such as fingerprint, iris, and gait recognition (Solanki & Pittalia, 2016). For example, in surveillance systems, instead of requiring people to place their hands on a reader (fingerprinting) or precisely positioning their eyes in front of a scanner (iris recognition), face recognition systems discreetly take pictures of individuals’ faces as they enter a specified area. There is no incursion or capture delay since the subjects are entirely oblivious of the actions, hence they (subjects) do not feel monitored under surveillance or their privacy invaded.
Expression recognition is a task that humans perform daily and effortlessly but that is not yet easily performed by computers without any challenge. These challenges include but are not limited to the different poses of facial images, partial occlusion, illumination variations, and facial expressions (e.g. facial features from one subject in two different expressions as expressed in fig. 1 below) which is why FE recognition remains a daunting task. In addition, certain expressions like ”sad” and ”fear” are in some cases very similar (Martinez & Valstar, 2016), which Darwin suggested manifests in animals as well (Darwin & Prodger, 1998). Hence, efficiently recognizing the different FEs is important for both the evaluation of emotional states and the automated face recognitions.
Figure 1. Similar Facial Expression being expressed by three different subjects. The images are from JAFFE and CK+ databases (Naidoo, Tapamo, & Khutlang, 2018).