Q1. How does the MCN Security facial biometric
technology handle false negatives?
A1. The MCN Security facial biometric system improves generality and accuracy (reduction in FAR/FRR) as additional enrollment images are captured and processed by the system. Procedurally, the application level systems are designed to allow enrollment at multiple stations (facility locations, access points or workstations) and provides the option for continual enrollment to compensate for effects of natural aging or other factors that lead to gradual change in appearance.
back to the top
Q2. How does the MCN Security facial biometric
technology increase a user's security level?
A2. In security applications,
facial biometrics are combined with either a secondary biometric (i.e. fingerprint,
eye scan) or a token such has an access code or proximity card. Facial biometrics
does not provide a 100% secure method of logical or physical access protection
on their own, however it can be used to reduce to risk of fraud through theft
of identity (access card/codes) by up to 99%. In the case of automated surveillance,
in which no secondary biometric may be practicable, 100% accuracy in not requirement
in that the object is generally to cue the attention of a live observer or
guard to a potential alarm/action scenario.
back to the top
Q3. How long does it take for the program
to make a decision (positive/negative)?
A3. The classification
rate (ie. rate of processing biometric templates is currently in the range
of 25,000-50,000 templates/second). The latency time for head tracking ranges
from sub-second to approximately 2 seconds depending upon distance from the
optical sensor (i.e. size of head relative to frame height) and the number
of individuals currently being tracked by the system. In a live scenario (i.e.
video based sensor) the decision latency is therefore primarily dependent
upon head tracking latency.
back to the top
Q4. Why is this process faster when there
are more images in the database
than fewer?
A4. Greater numbers
and variation of enrollment images tends to make recognition faster, however
more importantly, greater numbers of enrollment images extend the range of
head orientations, ambient lighting conditions and expression over which the
system (biometric template) will recognize the individual. These enrollment
images must of course incorporate these variations, which in most circumstances
may be captured applied a brief enrollment procedure that requires the user
to move (i.e. walking along a corridor, suggestion made to look left and right,
etc.)
back to the top
Q5. What are the limitation of this system…
facial hair, lighting,
facing/profiles, interference (e.g. in a crowded corridor with others passing
in front of field of vision, etc)?
A5. The limitations
of the systems are:
- The image capture must be compatible with optical wavelength sensor (not IR).
- The head orientation must be within +/- 90 degrees of full frontal and +/-30 degrees vertical.
- We must capture a minimum 15 pixels between eyes. The eyes, nose and mouth
must not be obscured.
Q6. What exactly is being measured/examined
to make an ID?
A6. We measure the face
in its entirety and we do not apply specific measurements of facial features.
The process more specifically involves:
- facial region is resampled to 64x64 pixels.
- FFT conversion of pixel image to frequency domain coefficients
- Quantization of FFT coefficients
- Above preprocessing operations supply target input to neural system that is then trained to discriminate between enrolled subject and general population
back to the top
Q7. What are the technological advantages
of this system over others?
A7. The technological advantage of this system over others are:
- It operates in very low resolution (down to 15 pixels between eyes).
- ) It provides all aspect facial recognition (up to +/- 90 degrees from
full frontal).
- It provides real time head acquisition in cluttered and moving environments.
- Facilitates continuous updating and refinement of biometric templates
through on-line learning or learning across multiple application stations
or sites.
- Extremely flexible application platform whereby tracking threads (up to 16) are controlled individually for operations such as enrollment, verification (one-to-one identification) and classification (one-to-many identification).
- To our knowledge, is the only product that gives you 7 primary operating
modes:
- Proximity card or access code only
- Proximity card/access code and facial recognition
- Proximity card/access code and fingerprint verification
- Facial verification only
- Fingerprint verification only
- Fingerprint and facial verification (dual factor biometric)
- Proximity card/access code with dual factor biometric
back to the top
HOME