There are continuous improvements being made in the field of facial recognition technology and similar AI-controlled biometric systems. However, this does not necessarily mean that the algorithms are unflawed or limited in any way. There are countless factors that are considered when the software recognizes and analyzes a face read from a digital camera’s input. The amount of light received by the camera versus the time of the day or how much sunlight is in the sky at that current time and at what angle might be details that have to be considered. The angle of the head structure of the person of interest and the direction in which his or her face is pointing might also be features that have to be acknowledged. It is also very interesting to note that some higher end facial recognition software have been able to analyze in real-time the velocity of the moving object (in this case, it is the person) and determine the angle at which the individual is facing as well as the degree of distortion of size in terms of the height, length, width, and circumference of the head and its individual structures as well as the body, in order to differentiate between individuals who look very similar as well as determine whether the image of the face was in fact an expressed emotion or an actual face of another person. The individual factors from the face that the software can measure can comprise of anything from the distance between the eyes, the length of the jaw, the width of the nose, the depth of the eye sockets, to the shape of the cheekbones (Bonsor and Johnson, 2011). These seemingly random but detailed factors
17 are all taken into account and importantly considered under a complex process that most, if not all, facial recognition softwares generally follow. The technology is designed to follow a fundamental algorithm that will collect data most likely in the form of an image from a real-time security camera to analyze, and either match the processed input to a preexisting identity in the systems continually growing database or create a new profile for the unidentified person. This traditional process followed by earlier developments in this technology is accomplished in following three simple steps: detection of the facial pattern, creation of the faceprint, and verification or identification of the profile(Bonsor & Johnson, 2011). Paradoxically, each individual step consists of its own complex procedures that take into consideration many parameters to ensure that the data provided from one step to another is as accurate as can be. Before the recognition software is released to the market or even deemed ready for use, however, the database is exposed to an initial set of faces for calibration. This is very similar to a beta version of software available to the public to be tested, but the actual facial recognition program is not released. It is merely a version to be certified by the developers. This facial template will allow the creator to see if the program is able to detect the face, and record the input properly into its database.