Landmarks of a face

Landmarks of a face

Facial Recognition Technology

Creator
Christoph von der Malsburg, graduate students of the University of Bochum in Germany, and the University of Southern California
Date(s)
1997
Facial recognition software utilizes automated computational analysis to recognize and then measure the various features of a face. Since every face has different distinguishable “landmarks,” or different shapes and contours that define facial features, these landmarks are defined as “nodal points." These nodal points are defined by many attributes, such as the width of the nose, depth of the eye sockets, shape of the cheekbones, length of the jaw line, distance between the eyes, and so on – there are about 80 nodal points on each human face. These nodal points are measured and these measurements become a numerical code, called a faceprint, which represents that specific face in the database (Bosner and Johnson 2001).

Before the recent popularization of 3D image technology, facial recognition software primarily relied on a comparison process between 2D images (“Facial Recognition”, 2006). There is a four-stage process as to how this comparison process takes place. First, the biometric system captures the “physical or behavior sample” during enrollment, then the system extracts unique date from the sample and a template is created. After the system begins its comparison process, the template is compared with a new sample, and finally the system decides if the extracted features from the new sample match or not (“Facial Recognition – Technology Overview”).

However, one can easily see the many flaws in this system – two evident flaws are 1. Unique faces are unaccounted for, 2. Any movement or shifted positioning would be accounted for (“How Does Facial Recognition Technology Work?” 2015). Clearly, this system becomes particularly succeptable to recognizing faces that are underpresented, or in other words "unique." Since the comparison process uses a set of sample images dictated by the technology developers themselves, often characteristics that don't reflect the developers, or targetted demographic, in turn, are unaccounted for (i.e. most often, POC). For instance in Figures 2, 3, and 4, these technologies' facial recognition software fail because they do not recgonize different skin colors or ethnicities.
Sources
"19 Cases Of Accidental Racism." Ebaumsworld. November 20, 2014. Accessed December 11, 2015. http://www.ebaumsworld.com/pictures/view/84364023/.
Bonsor, Kevin, and Ryan Johnson. "How Facial Recognition Systems Work." HowStuffWorks. N.p., 03 Sept. 2001. Web. 01 Dec. 2015. <http://electronics.howstuffworks.com/gadgets/high-tech-gadgets/facial-recognition1.htm>.
Charles, Craig. "Did You Know Facebook Can Identify YOUR Face." ThatsNonsense. February 9, 2015. Accessed December 11, 2015. http://www.thatsnonsense.com/know-facebook-can-identify-face/.
"Face Recognition." SpringerReference (2011): n. pag. Biometrics. National Science and Technology Council, 7 Aug. 2006. Web. 24 Nov. 2015. <http://www.biometrics.gov/Documents/FaceRec.pdf>.
"Face Recognition - Technology Overview." How Facial Recognition Works, Face Recognition, Facial Biometric Technology. Ex-Sight, n.d. Web. 02 Dec. 2015. <http://www.ex-sight.com/technology.htm>.
"How Does Facial Recognition Technology Work?" We Live Security. ESET, 24 Aug. 2015. Web. 01 Dec. 2015. <http://www.welivesecurity.com/2015/08/24/facial-recognition-technology-work/>.
Files