Biometric Concepts

Document Sample
Biometric Concepts Powered By Docstoc
					                                                          Biometric Concepts


                    2-D Face Recognition
                    A face recognition system that examines a single facial image, or a small set of facial images in
                    order to authenticate identity or find possible matches for the individual within a database of
                    images. Because 2-D systems use a smaller set of information to authenticate or determine
                    identity, such systems are inherently less reliable than 3-D systems. However, 2-D systems are
                    useful for watch list applications, and for any application in which only a limited number of
                    images are available.

                    3-D Face Recognition
                    A face recognition system that examines facial images within a live or taped video feed to view
                    multiple images of the face from multiple angles and compare the images to a biometric
                    template (or engram) in order to authenticate identity. AcSys FRS is a 3-D face recognition system.

                    See Automatic Banking Machine.

                    Access Control
                    A common use for biometric authentication. In an access control scenario, the biometric
                    authentication system is generally linked to a door-strike mechanism or other means of granting
                    access to a room or building. In verification mode, the user presents an identity token (e.g., a
                    proximity card) as well as a biometric sample (e.g., facial images, fingerprint, iris) to the system
                    and the system verifies the identity of the user by comparing the biometric sample against the
                    biometric template on record for the claimed identity. In identification mode, the user presents
                    only a biometric sample and the system compares that sample against all biometric templates on
                    record to find a match. Once the user is authenticated, the system activates the door-strike
                    mechanism to allow the user to pass. A common problem in access control systems is tailgating.
                    To completely eliminate tailgating, a mantrap door or other such solution must be implemented.

                    See Artificial Intelligence.

                    A sequence of instructions describing a step-by-step process used by a processing engine to
                    solve a particular problem.
 399 Pearl Street
  Burlington, ON    ANN
Canada L7R 2M8      See Artificial Neural Network.
P: 905-634-4477
F: 905-634-1101
                    Artificial Intelligence (AI)
                    The area of computer science that attempts to reproduce aspects of human thought and
                    cognition on computers. It can also be seen as an attempt to solve by computer any problem that
                    a human can currently solve more efficiently. An example of such a problem is the attempt to
                    design a machine to process human language. Such a machine might eventually be able to learn,
                    understand, and speak a human language as well as humans can.

                    Artificial Neural Network (ANN)
                    An artificial neural network applies artificial intelligence principles to allow a computer to learn
                    from experience.

                    See Automated Fingerprint Information System.

                    See Automatic Banking Machine.

                    The submission of a biometric sample to a biometric system for authentication.

                    The act of determining identity (in identification mode) or certifying identity (in verification
                    mode). Biometric authentication provides the ultimate in accountability, because it uses an
                    individual’s physical or behavioral traits to determine or certify identity. These traits are virtually
                    impossible for another individual to steal or mimic.

                    Automated Fingerprint Information System
                    A biometric system, associated primarily with criminal forensics, that compares a single finger
                    image with a database of finger images.

                    Automatic Banking Machine (ABM)
                    A machine that provides access to automated financial services. ABMs generally require a user to
                    insert a card and then key in a Personal Identification Number (PIN). The security of such
                    banking cards is questionable, as most PINs are specified by users and, because many people
                    have trouble remembering random numbers, they tend to specify PINs that can be guessed
                    easily. Other people use a random number, but then write the number down and place it in their
                    wallets. ABMs can be made secure through biometric authentication. In such a scenario, the
                    biometric sample essentially replaces the PIN.

                    Automatic Teller Machine (ATM)
                    See Automatic Banking Machine.

                    Behavioral Biometric
                    A behavioral trait used to authenticate identity. Behavioral biometrics can be learned
                    characteristics (e.g., keyboard dynamics). Contrast Physical Biometric.

                    A measurable (“metric”) characteristic, whether physiological or behavioral, of a living organism
                    (“bio”) that can be used to differentiate that organism as an individual.

                    Biometric Sample
                    Biometric data captured when the user makes an attempt to be authenticated by the system.
 399 Pearl Street
  Burlington, ON    Biometric Template
Canada L7R 2M8
P: 905-634-4477
                    A biometric representation of an individual. This data is used by the biometric system for real-
F: 905-634-1101
                    time comparison against biometric samples. The biometric template is created through the
                    enrollment process, in which initial biometric samples are captured.

                    The science and technology of recording and authenticating identity using physiological or
                    behavioral characteristics of the subject.

                    Imitating the processes of a living organism. AcSys FRS is referred to as biomimetic, because it
                    emulates the processes of the human brain at the cellular level. Applied biomimetic intelligence
                    allows AcSys FRS to learn and recall in much the same way a human does, and to augment
                    existing memories (i.e., biometric templates or engrams) with new memories without erasing the
                    original memory or increasing storage requirements for the compounded memory.

                    The act of collecting a biometric sample (e.g., facial images).

                    Learning. Cognitive science studies the ability of the brain to learn. Recognition (re-cognition) is
                    the ability of the brain to recall what has been previously learned.

                    Being in two places at once. Co-location houses provide secure data storage services, usually for
                    backup or mirroring purposes.

                    The process of comparing a biometric sample with a previously stored biometric template (one-
                    to-one) or templates (one-to-many).

                    The face recognition method developed in the late 1980s by scientists at MIT. Also known as
                    Principal Component Analysis (PCA). The word “eigenface”, half German and half English, can
                    be translated roughly as “one’s own face”. In terms of a technical definition, an eigenface is a
                    two-dimensional, grayscale image representing distinctive characteristics of a facial image. The
                    distinctive features are stored by the system as coefficients.

                    Advocates of this technology maintain that most human faces can be reconstructed by combining
                    features drawn from a set of between 100 and 150 such eigenfaces. In other words, although
                    there are more than 5 billion individual human faces on this planet, they can all be
                    reconstructed from fewer than 150 eigenfaces. Eigenface components are used to build facial
                    images. Software applications using this technology analyze a facial image, calculate its
                    eigenface value using distinctive feature coefficients (essentially deconstructing the face into its
                    eigenface components), and compare that value to a biometric template or database of biometric
                    templates to find matching or similar eigenface values.

                    The ability of systems using this technology to match faces to biometric templates is highly
                    dependent on lighting and facial angle. Since the biometric templates generally do not
                    incorporate a variety of images taken in different lighting conditions and at different angles,
                    eigenface technology works best when the system views faces frontally and in consistent lighting
                    conditions. Its ability to generalize across facial poses has been found to be rather poor.

 399 Pearl Street   Engram
  Burlington, ON    Literally, a memory trace. A biometric template is referred to as an engram in AcSys FRS
Canada L7R 2M8      documentation to reflect similarities with neurophysiology and the cognitive processes of the
P: 905-634-4477
                    brain, particularly the brain's ability to enfold multiple memories on a single storage space.
F: 905-634-1101
                    AcSys FRS emulates human cognition by enfolding multiple memories of an individual onto a
                    single biometric template without increasing the size of the template. See also Biomimetic.

                    Before a biometric system can differentiate an individual, it must have a biometric template, of
                    that individual. The act of capturing biometric samples of an individual and creating the
                    biometric template to represent that individual is known as enrollment.

                    Face-in-a-Crowd Surveillance
                    The act of capturing facial images of subjects without their cooperation and comparing those
                    images against a watch list. Compare to authentication, in which an individual seeks recognition
                    by the face recognition system. Face-in-a-crowd surveillance gets far more media attention than
                    authentication applications. Much of that attention is negative, "Big Brother" type coverage,
                    which focuses on potential abuses and perceived invasions of privacy. The ethics and legalities of
                    surveillance and privacy protection are important issues to be addressed with regard to face-in-a-
                    crowd applications.

                    Another aspect of face-in-a-crowd surveillance frequently reported in the media is the apparent
                    failure of high-profile implementations to detect suspected criminals while misidentifying
                    innocent people. These failures can largely be attributed to failures in tracking and

                    Facial Metrics
                    See Local Feature Analysis.

                    Failure to Enroll (FTE)
                    The rate at which a system fails to capture and process initial biometric samples needed to build
                    a biometric template for an individual.

                    False Acceptance Rate (FAR)
                    The rate at which a biometric system incorrectly authenticates an identity or fails to reject an

                    False Rejection Rate (FRR)
                    The rate at which a biometric system fails to authenticate the identity of a legitimate user.

                    See False Acceptance Rate.

                    See False Rejection Rate.

                    See Failure to Enroll.

                    Generalization refers to the ability of a biometric system to receive and process a biometric
                    sample in less than optimal conditions and find a matching biometric template -- in other words,
                    its ability to move from the specific representation (the biometric sample) to the general
                    representation (the biometric template).

 399 Pearl Street   See Holographic/Quantum Neural Technology.
  Burlington, ON
Canada L7R 2M8
P: 905-634-4477
F: 905-634-1101
                    Holographic/Quantum Neural Technology (HNeT)
                    AND Corporation's proprietary core technology which allows AcSys FRS to emulate human
                    cognition. AcSys Biometrics Corp. holds the worldwide rights to HNeT for face and speech
                    recognition applications. See also Engram, Biomimetic.

                    Authentication by identification occurs when the individual provides only biometric input (a
                    biometric sample) and the system compares that biometric input against all biometric templates
                    on record to determine the individual's identity. With AcSys FRS, an individual steps in front of a
                    camera and the system captures multiple images. The system then compares this biometric
                    sample against the biometric templates in the database and eliminates templates until it is
                    certain of a match. When the system matches the sample to a record, the system identifies the
                    individual as the person represented by the database record. Systems that identify in this manner
                    are said to use "one-to-many" face recognition, because they compare one individual to many
                    biometric templates in order to determine the user's identity. Contrast Verification.

                    A person who submits a biometric sample in either an intentional or inadvertent attempt to pass
                    himself or herself off as another person who is an enrollee.

                    See Local Feature Analysis.

                    Local Feature Analysis (LFA)
                    In the early 1990s, researchers at Rockefeller University developed a face recognition technology
                    that came to be known as Local Feature Analysis (LFA). This technology extracts a set of local
                    features from a set of sample human faces and uses them as building elements. It then
                    combines these elements, as well as their relative arrangement, as a means of verifying and/or
                    identifying faces.

                    This technology is similar to Eigenface technology, but is said by its proponents to be less
                    sensitive to facial expressions and changes in pose and lighting. It measures a face according to
                    a subset of nodal points within the inner region of the face (e.g. the distance between the inside
                    corners of the eyes, the distance between the outside corners of the eyes and the outside corners
                    of the mouth) and the relationship between these measurements. However, it still takes a
                    mechanical approach to the problem of face recognition, differentiating faces according to a
                    small set of measurements and relations between points. Contrast Biomimetic.

                    Mantrap Door
                    Actually a set of two doors, of which only one can be open at any given time. By installing linked
                    biometric access control systems at each door, a mantrap door system can eliminate tailgating.

                    The process of comparing a biometric sample against a previously stored biometric template and
                    scoring the level of similarity. An accept or reject decision is then based upon whether this score
                    exceeds a pre-established threshold.

                    Describes the technique of comparing a biometric sample against all biometric templates in a
                    database to determine identity. The one-to-many technique is used in identification.

 399 Pearl Street   One-to-One
  Burlington, ON
                    Describes the technique of comparing a biometric sample to a single biometric template to
Canada L7R 2M8
P: 905-634-4477     certify identity. The one-to-one technique is used in verification.
F: 905-634-1101
                    See Eigenface.

                    Physical Biometric
                    A physical characteristic used to authenticate or determine identity. Contrast Behavioral

                    Principal Component Analysis
                    See Eigenface.

                    The right of an individual to keep information about his (or her) life from the knowledge and
                    attention of others, including government organizations and commercial enterprises, and to
                    remain free from outside intervention except under the provisions of law. (From ODLIS)

                    See also Identity and Face-in-a-Crowd Surveillance.

                    A cognitive disorder in which the afflicted person is unable to recognize faces. Victims of
                    prosopagnosis must use other cues, such as voice, to identify individuals, even if they have
                    known those individuals all of their lives. Prosopagnosis demonstrates the primacy of face
                    recognition among the human cognitive abilities used to identify other people.

                    Real-time Learning
                    The ability of a system to learn immediately from experience.

                    See Face-in-a-Crowd Surveillance.

                    The act of illicitly following an authenticated user through a door without being authenticated
                    oneself. To completely eliminate tailgating, a mantrap door system must be implemented. Other
                    solutions such as optical tailgating technology or turnstiles can be used to detect tailgating, but
                    they cannot eliminate it completely.

                    Time & Attendance
                    A popular application for biometric authentication. Using a biometric authentication application
                    in such a scenario prevents individuals from logging each other in and out of a facility or system.

                    Tracking is a measurement of a face recognition system's ability to recognize facial patterns
                    within an image, and to follow those patterns as they move within the field of view of the camera.

                    Verification occurs when the individual provides both biometric and non-biometric input and the
                    system verifies the biometric evidence against the non-biometric. With a face recognition system,
                    an individual might swipe a card that represents him as "John Smith". The individual then stands
                    in front of a video camera and the system captures images and compares them against the
                    biometric template of John Smith on record. If the individual is indeed the John Smith on
                    record, the system verifies his identity. This verification scenario is also known as "one-to-one"
                    face recognition, because the system matches one user to one database record. Contrast
 399 Pearl Street
  Burlington, ON
Canada L7R 2M8
P: 905-634-4477
F: 905-634-1101
                    Watch List
                    A set of biometric samples or biometric templates against which newly captured biometric
                    samples are compared for resemblance. Watch lists are most commonly used with face
                    recognition technology, because face recognition is the only biometric technology that can
                    capture biometric samples at a distance (i.e., without the subject being aware of the process).
                    Such watch list applications generally use a 2-D approach to face recognition: they find
                    resemblance using only a single image or a small set of images, rather than examining the whole
                    face from a variety of angles (3-D). The 2-D approach is limited in its ability to determine
                    identity, so such applications generally return a list of possible matches rather than a single
                    identity. If any of the possible matches score above a predetermined threshold, the appropriate
                    security officers are alerted.

 399 Pearl Street
  Burlington, ON
Canada L7R 2M8
P: 905-634-4477
F: 905-634-1101