A facial occlusion position is one of a number of possible poses in which something, such as a person's hand, blocks (occludes) part of their face. The occlusion may result from one or both hands either directly on the face or in front of it from the viewer's perspective. Other common examples of objects responsible for facial occlusion are hair, hats and sunglasses.
Positions in which hands partially cover faces are a difficult problem for facial recognition and object recognition software because people's hands have many characteristics in common with their faces, such as coloring and texture. As a result, images in which people's hands hide part of their faces are often rejected by the software because they are considered to contain too much noise to be useful.
However, such images could potentially yield valuable information because people use their hands in conjunction with their faces to communicate through gestures. One example is the common facepalm position used to convey a range of emotions including frustration, embarrassment, distress, shock and disbelief.
In a similar but distinct position, people often lean forward resting the forehead on a hand or fist to express thoughtfulness. A palm slapping on the forehead can be interpreted as a sudden recollection of something that had been forgotten or a "lightbulb" moment when something becomes immediately clear. The back of the hand on the forehead, in contrast, expresses melodramatic distress. Resting the chin on one or both hands may express interest if the eyes are engaged or thoughtfulness if the individual is looking away from the viewer.
Current research is exploring methods of creating reconstructions of images with and without occlusion. Another line of study is focused on interpreting various facial occlusion poses, standardizing and encoding that information. The interpretation of expressions, gestures and other non-verbal cues is important to an increasing number of fields including software design, marketing, robotics and artificial intelligence (AI).