Transcript Kein Folientitel

Peter Zukowski
Biometry
Peter Zukowski
Biometry
Seminar of Multimedia
Peter Zukowski
Biometry
Why biometry?
Three methods of identification:
 knowledge (e.g. password, PIN)
 possession (e.g. key, card)
 biometry (physiological characteristics)
Biometric characteristics
 cannot be lost, copied, stolen, etc.
 need not to be kept, remembered, etc.
 are naturally fixed to the user
Peter Zukowski
Biometry
Contents
Biological background
 DNA and genetic code
 Reproduction
 Inheritance and gene expression
Biometry in general
 Biometric characteristics
 Biometric systems
 Biometric standards and interfaces
Biometric techniques
 Voice recognition
 Iris recognition
 Fingerprint recognition
Peter Zukowski
Biometry
DNA
Everyone of us has got approx. 1014 cells.
Each of them (with some exceptions) contains our whole
genetic information called the “genome”.
This information is encoded in 4 kinds of nucleobases:
A
Adenine
G
Guanine
C
Cytosine
T
Thymine
Sequence of these bases  genetic code.
Connected to a chain: the DNA
DNA is located on chromosomes in the cell core.
Peter Zukowski
Biometry
DNA
Nucleobases are organized in nucleotides:
Nucleotide = phosphate + deoxyribose + nucleobase
DNA = deoxyribonucletic acid
P
P
P
P
P
P
D
D
D
D
D
D
A
G
T
A
C
T
T
C
A
T
G
A
D
D
D
D
D
D
P
P
P
P
P
DNA chain
DNA double helix
DNA chain
P
Peter Zukowski
Biometry
Genetic code
enzyme
hormone
chain
hemoglobin
amino acid
protein
encodes
A
encodes
G
T
A
codeword of genetic code.
G
A
C
T
T
gene
A
C
T
G
G
C
A
Peter Zukowski
Biometry
Chromosomes
Whole genetic information  spread over 23 chromosomes:
Normal cell:
two sets of
chromosomes
 23 from mother
 23 from father
Peter Zukowski
Biometry
Reproduction
46x
46x
cell
meiosis
23x
meiosis
23x
ovule
fertilization
46x
zygote
mitosis
46x
cell
embryo
sperm cell
Peter Zukowski
Biometry
Twins
Fraternal twins:
Two fertilizations
at the same time
 2 different genomes  no problem with biometrics
Identical twins:
cloning of whole
embryonic complex
 2 identical genomes  problem?
Peter Zukowski
Biometry
Gene expression
How will a gene be expressed? (And which gene?)
Environmental factors (e.g. lack of a certain substance)
Repetitive genes
 many genes that code one characteristic
 May vary from each other
 Which will be expressed is a matter of accident
?
genes
protein
?
Peter Zukowski
Biometry
Genes and biometry: Example
Finger pattern:
Consists of many proteins,
influenced by repetitive genes.
 Many components: Connective tissue filaments,
blood vessels, sweat glands, receptors, skin covering layer
 Unique pattern aroused by factors influencing each other
Suitable for biometric identification even for identical twins !
Blood type:
One gene for A and one gene for B.
If both genes exist in a genome  type AB, if none  type 0
Not suitable for biometric identification (e.g. identical twins)
Peter Zukowski
Biometry
Conclusion for biometry
Genotypic characteristics:
 Inheritance / genetic code
Randotypic characteristics:
 gene expression during development.
Mutual influence of genes enables unique combinations.
Applicable even for identical twins (and clones) !
Peter Zukowski
Biometry
Contents
Biological background
 DNA and genetic code
 Reproduction
 Inheritance and gene expression
Biometry in general
 biometric characteristics
 biometric systems
 biometric standards and interfaces
Biometric processes
 Voice recognition
 Iris recognition
 Fingerprint recognition
Peter Zukowski
Biometry
Characteristics used in biometry
Finger pattern
Image and pattern of finger surface
Hand geometry
Measurements and shape of hand and fingers
Face geometry
Image of the face and components
Iris
Pattern of the tissue around the pupil
Retina
Pattern of blood vessels in background of eye
DNA
Coding of DNA as carrier of the human genome
Speech
Tone and speaking behavior
Handwriting and
signature
Pressure, speed, acceleration, style
Typing rhythm
Dynamics of keystrokes on keyboard
Mimics
Voice tone and speech behavior
Lips
Movements of the lips
Palm
Image on surface of hand
Vein
Structure of veins on arm
Peter Zukowski
Biometry
Characteristics used in biometry
Which characteristics are suitable for biometry?
Criteria:
Uniqueness: Is it unique in every individual?
Universality: Is it available in every individual?
Constance: Does it change in time?
Applicability: Is it measurable?
Comfort: Is it user-friendly?
Peter Zukowski
Biometry
Characteristics used in biometry
Which characteristics are most suitable?
Characteristics
Uniqueness
Universality
Constance
Applicability
Comfort
Finger pattern
100%
100%
80%
80%
80%
Hand geometry
80%
100%
80%
60%
80%
Face geometry
80%
100%
60%
60%
100%
100%
100%
100%
80%
90%
Retina
90%
100%
80%
60%
70%
DNA
90%
100%
90%
50%
80%
Speech, tone
50%
80%
60%
70%
50%
Handwriting
50%
60%
40%
60%
40%
Typing rhythm
30%
40%
40%
60%
50%
Mimics
60%
100%
60%
30%
40%
Iris
Peter Zukowski
Biometry
Hardware examples
Hand geometry scanner:
Retina camera:
Handwriting pad:
Peter Zukowski
Biometry
Identification systems
User registers himself: “enrollment” process
User comes back for “identification” or “verification”
Identification: User recognized out of many individuals.
1:n
 “Who is he/she ?”
Verification: A requested identification is accepted or rejected.
1:1
 “Is it him/her ?”
Peter Zukowski
Biometry
Enrollment
Measurement of chosen characteristics.
Extraction of characteristics (signal processing).
Eventually compression and encoding of data set.
Storing of data set as a template.
Verification: Only one template needed !
Identification: All templates have to be available !
 Problems: Data management, distribution, storage,
transmission, protection, security, legal problems
Peter Zukowski
Biometry
Identification
Authorized person must not be rejected.
 False reject rate
Unauthorized person must not be accepted.
 False accept rate
Usually data sets are classified (faster search)
Never an exact matching is requested.
 Matching within a certain tolerance field !
Choice of tolerance rates and threshold values is
major problem
 training and adjustment of system
 by developer (e.g. iris) or/and by user (e.g. speech)
Peter Zukowski
Biometry
Application fields
Forensics (e.g. AFIS, FBI)
Access security (buildings, etc.)
National security (e.g. US-VISIT, US home dep. of security)
Staff identification in company
Log-in / Log-out
Personalization (e.g. profiles in computer)
Protection of devices (e.g. memory sticks)
Smart cards
…
Peter Zukowski
Biometry
Standardization and APIs
BioAPI
 Bio API Consortium
Human Authentication API (HA-API)
 US-department of defense
Speaker Verification API (SVAPI)
 SRAPI Consortium
Common Biometric Exchange File Format (CBEFF)
 Novell, NIST (Nat. Institute of Std. and Technol.)
Peter Zukowski
Biometry
Contents
Biological background
 DNA and genetic code
 Reproduction
 Inheritance and gene expression
Biometry in general
 biometric characteristics
 biometric systems
 biometric standards and interfaces
Biometric processes
 Voice recognition
 Iris recognition
 Fingerprint recognition
Peter Zukowski
Biometry
Why speaker recognition?
Cheapest and most spread technique.
Not very suitable for security applications but
dominating on consumer market
 boom of speech recognition software.
Speech recognition:
 text recognition, dictation systems, input of instructions.
Speaker recognition:
 Speech rhythm and tone of voice.
Text-dependent or text-independent.
Disadvantage: time-consuming enrollment.
Peter Zukowski
Biometry
Speaker recognition
Orientation towards functionality of human speech:
White noise
Acoustic filter
Cavities in mouth / nose form a tone
Glottis creates oscillation with certain pitch
Peter Zukowski
Biometry
Speaker recognition
Recorded signal is divided into many time samples
(10 – 30 ms).
The signal is transformed into frequency domain.
For each of the samples a digital filter is computed.
A filter is described by e.g. 5 parameters.
Training (enrollment):
 Storage of many templates for comparison.
 Storage of only one template but repetitive optimization.
Peter Zukowski
Biometry
Why iris recognition?
Most reliable, exact and secure technique.
Unfortunately most expensive technique.
The iris the most unique and constant characteristic.
Enrollment only once in a lifetime.
Well protected by eye lids.
Very quick enrollment (approx. 2 minutes)
Not detrimental to health.
Peter Zukowski
Biometry
Iris recognition
Digital monochrome camera.
Iris extracted from rest.
All computations: Relatively to radius (r=1)
 size does not matter
Location and orientation: reversible operations.
Phase information is extracted using wavelets.
Phase
information
encoded in
complex
numbers:
Peter Zukowski
Biometry
Why fingerprint recognition?
Fingerprints are unique even for identical twins.
Normally this method can be applicable to every man.
A fingerprint never changes during a lifetime
(if no injuries, etc.).
Very old and experienced technique.
Criminalistics: few hundred millions of fingerprints taken.
Peter Zukowski
Biometry
Characteristics of fingerprints
Made of a series of ridges and furrows.
Pattern type: used for classification (faster search).
loops
bows
whirls
(60%)
(5%)
(35%)
Minutiae points (ridge terminations, ridge bifurcations)
Whole image of fingerprint: only used in forensics (250 kB).
In other applications data set consists of minutiae (< 1 kB).
Peter Zukowski
Biometry
Characteristics of fingerprints
Minutiae types:
Rod
Ellipsis
Whorl
Bifurcation
Tented
arch
Loop
Island
Arch
Sample of a minutiae extraction:
Ridge termination
Ridge bifurcation
Peter Zukowski
Biometry
Optical sensors
Finger is placed on transparent area in front of camera.
Advantages:
 Low costs
 Good resolution (up to 500dpi).
 No electrostatic and thermal detraction.
Disadvantages:
 Damage or soiling of contact area.
 Leaving of traces on contact area.
Example: touch less sensor by TST
Peter Zukowski
Biometry
Capacitive sensors
Finger and array of tiny condensers create electric field.
Pattern on finger detected through distance to skin.
Advantages:
 low costs, leading technique on consumer market.
 good durability
Disadvantages:
 traces on contact area  occurrence of “ghost-images”
Peter Zukowski
Biometry
Other sensor techniques
Thermal sensors: detect temperature differences on skin.
 difficult operation but not susceptible to latent traces
Ultrasound sensors: detect reflection characteristics on skin.
 Very complex but also very secure
Life checks (detection of pulse):
 acoustic (sound of heartbeat)
 capturing of motion of finger image
 ox metric pulse detection (as used in ambulances):
 detection of oxygen level in blood through light rays
Peter Zukowski
Biometry
Capacitive Sensors: Hardware
Siemens ID Mouse:
 Costs: approx. € 120
ARP Datacon Magic Finger USB Stick:
128 MB flash memory
 Costs: € 170
Smart Card:
Peter Zukowski
Biometry
Conclusions
Very significant and interesting topic.
Very sensitive field, legally and ethically.
Vast application field.
Much to do for a computer scientist.
Many companies from Germany.
Much to do in standardization.
Still developing but growing quickly.
There is more yet to come …
Peter Zukowski
Biometry
Biometry
The End
Thank You for Your attention !