Li S.Z., Jain A.K. (eds.) Encyclopedia of Biometrics

Подождите немного. Документ загружается.

Photography for Face Image Data. Figure 5 Example of a photo booth(2).

Photography for Face Image Data. Figure 6 Light and camera arrangement for a photo booth.

1080

Photography for Face Image Data

specular reﬂections on the face, and deep shadows on

background (which can blend with thesubject’s dark

hair, etc.), improve the accuracy of automated face rec-

ognition systems. Other factors that can affect face image

quality include pose, expression, and positioning. In-

creasingly, face image quality assessment software is

being used to ensure high quality face images, for the

preparation of not only e- , but also other ID documents.

Related Entries

▶ Enrollment

▶ Facial authentication

▶ Facial identiﬁcation

▶ Facial veriﬁcation

References

1. ‘‘Machine Readable Travel Documents, Part 1, Machine Read-

able Passports,’’ ICAO Doc9303 6th edn. 3rd draft, 18, Nov.

(2004)

2. FRVT2002. http://www.frvt.org/FRVT2002/default.htm

3. Phillips, P.J., et al.: FRVT 2006 and ICE 2006 Large-Scale Results,

NISTIR 7408, National Institute of Standards and Technology,

Gaithersburg, MD 20899, USA. http://face.nist.gov/frvt/frvt2006/

frvt2006.htm 9. Guidelines of USA, UK, Australia and Canada

4. ISO/IEC 19794-5:2005, Information Technology – Biometric

Data Interchange Formats – Part 5: Face Image Data (2005)

5. http://travel.state.gov/pdf/Photo%20Guide%2010-01-04.pdf

6. http://www.ips.gov.uk/passport/downloads/Photographers-guide-

to-passport-photos-edited.pdf

7. https://www.passports.gov.au/Web/Requirements/Photos.aspx

8. http://www.ppt.gc.ca/cdn/photos.aspx; http://www.ppt.gc.ca/

form/pdfs/pptc320_eng.pdf

9. ISO/IEC 19794-5:2005, Information technology – Biometric data

interchange formats – Part 5: Face image data – AMENDMENT 1:

Conditions for taking photographs for face image data (2007)

10. Commissioned Report: Studies on Biometric Authentication

Using IC Passports. http://www.nmda.or.jp/nmda/bio/ic-pass

ports/ic-passports.pdf

11. Junichi Sakaki, Shizuo Sakamoto, Koji Matsunaka et al.: [Special

Talk] IC Passports and Systematic Evaluation toward Employing

Face Authentication Technologies, IEICE (The Institute of Elec-

tronics, Information and Communication Engineers) technical

report PRMU2005-92 (2005–10), pp. 51–56, (2005)

12. D’Amato , D .P.: Best Practices for Taking Face Photographs

and Face Image Quality Metrics NIST Biometric Quality Work-

shop. http://www.itl.nist.gov/iad/894.03/quality/workshop/proc/

d’amato_nist_image_quality_workshop_slides-ﬁnal.pdf Accessed

8 March (2006)

Photography Guidelines

▶ Photography for Face Image Data

Photography for Face Image Data. Figure 7 User interface: Head positioning frame.

Photography for Face Image Data

1081

Photometric Guidelines

▶ Photography for Face Image Data

Photon

In classical electromagnetic theory, light and other

forms of electromagnetic radiation propagate as

waves. Quantum mechanics teaches us that nature is

more complicated; light propagation can be modeled

as either a wave or a particle, depending on the view-

point and measurement process of the observer. A

photon is a ‘‘particle’’ of light, in the same sense that

an electron is a ‘‘particle’’ of electricity. When light is

absorbed by a piece of silicon in an camera sensor, the

model for the interaction is that photons of light are

individually absorbed by the silicon and excite elec-

tron-hole pairs in the silicon that result in an electrical

signal that can be processed to generate images.

▶ Iris Device

Physical Analogies for Ear

Recognition

DAV I D J. HURLEY,MARK S. NIXON

School of Electronics and Computer Science,

University of Southampton, Southampton, UK

Synonyms

Convergence feature extraction: convergence

Definition

In the context of ear biometrics, Hurley et al. [1–3] have

developed a pair of invertible linear transforms called

the

▶ force ﬁeld transform and the ▶ potential energy

transform, which transform an image into a force ﬁeld

by pretending that pixels have a mutual attraction

directly proportional to their intensities and inversely

proportional to the square of the distance between

them, rather like Newton’s Law of Universal Gravita-

tion. Underlying this force ﬁeld, there is an associated

potential energy ﬁeld, which in the case of an ear, takes

the form of a smooth surface with a number of peaks

joined by ridges. The peaks correspond to the potential

energy wells, and to extend the analogy the ridges

correspond to the potential energy channels. Since

the transform also happens to be invertible, all the

original information is preserved, and since the other-

wise smooth surface is modulated by these peaks and

ridges it is argued that much of the infor mation is

transferred to these features and that therefore they

should make good features for recognition. An analysis

of the mechanism of this algorithmic

▶ ﬁeld line fea-

ture extraction approach leads to a more powerful

method called

▶ convergence feature extraction,

based on the divergence of force direction revealing

even more features in the form of antiwells and

antichannels.

Introduction

The last 10 years has seen increasing interest in the ear

as a biometr ic with signiﬁcant contributions from

computer vision researchers [1–8]. In this context the

force ﬁeld transform has been developed that effective-

ly ﬁlters an ear image by convolving it with a huge

inverse square kernel more than four times the size of

the image, the force then being the gradient of the

resulting massively smoothed image. Force ﬁeld fea-

ture extraction subsequently exploits the directional

properties of the force ﬁeld to automatically locate

ear features in the form of potential channels and

wells. The force ﬁeld paradigm allows us to draw

upon a wealth of proven techniques from vector ﬁeld

calculus; for example, the divergence operator is ap-

plied to the force ﬁeld direction, yielding a nonlinear

operator called convergence of force direction leading

to the even more powerful convergence feature extrac-

tion. The extreme kernel size results in the smoothened

image having a general dome shape, which gives rise to

brightness sensitivity issues. However, it is argued by

showing that the ﬁeld line features are hardly distorted,

that this will have little overall effect, and this conclusion

is borne out by including brightness variation in the

1082

Photometric Guidelines

recognition tests. On the other hand, the dome shape

leads to an automatic extraction advantage, and this is

demonstrated by deliberately using poorly registered and

poorly extracted images in recognition tests and then

comparing the results with those for principal compo-

nents analysis (PCA) under the same conditions, where

weseethattheearimageshavetobeaccuratelyextracted

and registered for PCA to achieve comparable results.

The technique is validated by achieving a recognition

rate of 99.2% on a set of 252 ear images taken from the

XM2VTS face database [9]. Not only is the inherent

automatic extraction advantage demonstrated but it is

also shown that it performs even more favorably

against PCA under variable brightness conditions,

and also demonstrates its excellent noise per formance

by showing that noise has little effect on recognition

results. Thus the technique has been validated by

achieving good ear recognition results, and in the

process a contribution has been made to the mounting

evidence that the human ear has considerable biomet-

ric value.

Ear Feature Extraction

Force Field Feature Extraction

Here, the force ﬁeld transform and algorithmic ﬁeld

line feature extraction are described before intro-

ducing convergence feature extraction. The mathe-

matical concepts used can be found in basic works

on electromagnetics [10] and a more detailed descrip-

tion of the transform can be found in [3]. Faster

computation using convolution and the Fast Fourier

Transform (FFT) and the question of brightness sensi-

tivity, both theoretically and by demonstration, are

considered.

The image is ﬁrst transformed into a force ﬁeld

by treating the pixels as an array of mutually attract-

ing particles that attract each other according to the

product of their intensities and inversely to the square

of the distances between them. Each pixel is assumed

to generate a spherically sy mmetrical force ﬁeld so

that the total force Fðr

Þ exerted on a pixel of unit

intensity at the pixel location with position vector r

by remote pixels with position vector r

and pixel

intensities Pðr

Þ is given by the vector summation,

Fðr

Þ¼

Pðr

r

8i 6¼ j

08i ¼ j



: ð1Þ

The underlying energy ﬁeld Eðr

Þ is similarly des-

cribed by,

Eðr

Þ¼

Pðr

r

8i 6¼ j

08i ¼ j



: ð2Þ

In order to calculate the force and energy ﬁelds for

the entire image, the calculations should be performed

for every pixel. However, this requires the number of

applicati ons of eq uations 1 and 2 to be propor tional to

the square of the number of pixels. Therefore, for faster

calculation, the process is treated as a convolution of

the image with the force ﬁeld corresponding to a unit

value test pixel, and then invoking the Convolution

Theorem to perform the calculation as a frequency

domain multiplication. The result of this is then trans-

formed back into the spatial domain. The force ﬁeld

equation for an n-pixel image becomes,

force field ¼

ﬃﬃﬃ

 ℑ

1

½ℑðunit force fieldÞℑðimageÞ;

ð3Þ

where ℑ stands for the Fourier Transform and ℑ

1

for

its inverse. Listing 1

shows how to implement this in Mathcad in which 1j

denotes the complex operator and cfft and icfft denote

the Fourier and inverse Fourier transforms, respective-

ly. Moreover, because the technique is based on a

natural force ﬁeld there is the prospect of a hardware

implementation in silicon by mapping the image pixels

to electric charges, which would lead to very fast real

time force ﬁeld calculation.

Figure 1a demonstrates ﬁeld line feature extraction

for an ear image where a set of 44 test pixels is arranged

around the perimeter of the image and allowed to

Physical Analogies for Ear Recognition

1083

follow the ﬁeld direction so that their trajectories form

ﬁeld lines which capture the general ﬂow of the force

ﬁeld. The test pixel positions are advanced in incre-

ments of one pixel width and the test pixel locations

are maintained as real numbers, producing a smoother

trajectory than if they were constrained to occupy

exact pixel grid locations. Note the two obvious poten-

tial wells in the lower part of the ﬁeld.

Convergence Feature Extraction

This analytical method came about as a result of analyz-

ing in detail the mechanism of ﬁeld line feature extrac-

tion. As shown in Fig. 1d, when the arro ws normally used

to depict a force ﬁeld are replaced with unit magnitude

arrows, thus modeling the directional behavior of explor-

atory test pixels, it becomes appar ent that channels and

wells arise as a result of patterns of arrows con ve rging

toward each other , at the interfaces between regions of

almost uniform force direction. As this brings to mind the

divergence operator of vector calculus, it was natural to

inv estigate the nature of any relationship that might exist

between channels and wells and this operator. This

resulted not only in the disc ov ery of a close correspon-

dence between the two, but also rev ealed extra information

corresponding to the interfaces between diverging arro ws,

leading to a more general description of channels and wells

in the form of a mathematical function in which wells and

channels are revealed to be peaks and ridges, respectively,

in the function value. The new function maps the forc e

ﬁeld to a scalar ﬁeld, taking the force as input and return-

ing the additive inverse of the divergence of the force

direction. The function will be referred to as the

▶ force

direction convergence ﬁeld C( r) or just

▶ convergence

for brevity. A more formal deﬁnition is given

cðrÞ¼div fðrÞ¼ lim

DA!0

fðrÞdl

¼r:fðrÞ¼



;

ð4Þ

where fðrÞ¼FðrÞ

jFðrÞj, DA is incremental area, and

dl is its boundary outward normal. This func tion is

real valued and takes negative values as well as positive

ones where negative values correspond to force direc-

tion divergence. Listing 2

shows a particular implementation of convergence in

Mathcad where FF represents the force ﬁeld and DF is

the direction ﬁeld.

It must also be stressed that convergence is non lin-

ear because it is based on force direction rather than

force. This nonlinearity means that we are obliged to

Physical Analogies for Ear Recognition. Figure 1 Convergence field: (a) ﬁeld lines, (b) convergence ﬁeld, (c)

superimposition, and (d) force direction.

1084

Physical Analogies for Ear Recognition

perform the operations in the order shown; we cannot

take the divergence of the force and then divide by the

force magnitude. Div(grad/|grad|) 6¼ (div grad)/|grad|.

This is quite easily illustrated by a simple example

using the scalar ﬁeld e

in equa tion 5,

divðgrad=jgradjÞ

r



¼r

¼ri ¼ 0

()

6¼

ðdiv gradÞ=jgradj

rre

jre

¼ 1

()

;

ð5Þ

where i is a unit vector in the x direction. This illus-

trates that even though convergence looks very much

like a Laplacian operator, it deﬁnitely is not.

Figure 1 shows the relationship between ﬁeld lines

(a) and con v ergenc e (b) by merging the two ﬁelds in (c).

A small rectangular section of the force direction ﬁeld

indicated by a small rectangular insert in Fig. 1a and b is

shown magniﬁed in Fig. 1(d). We can see clearly that

channels coincide with white convergence ridges and

also that wells coincide with convergence peaks, which

appear as bright spots. Note the extra information in

thecenteroftheconvergencemapthatisnotinthe

ﬁeld line map. Negative conv erge nce values represe nting

antichannels appear as dark bands, and positive values

corresponding to channels appear as white bands. We see

that the antichannels are dominated by the channels, and

that the antichannels tend to lie within the conﬁnes of the

channels. Note also the correspondence between con-

verging arrows and white ridges, and between diverging

arrows and black ridges. The features detected tend to

form in the center of the ﬁeld due to its ov erall dome

shape, with channels and wells tending to follow intensity

ridges and peaks while antichannels and antiwells tend to

follow intensity troughs and hollows.

Brightness Change Analysis

Before proceeding to the next section on ear recogni-

tion, the effect of brightness change will ﬁrst be ana-

lyzed by considering its effect on the energy ﬁeld and

then conﬁrmed by visual experiment. Should the indi-

vidual pixel intensity be scaled by a factor a and

also have an additive intensity component b,we

would have

Eðr

Þ¼

aPðr

Þþb

r

8i 6¼ j

08i ¼ j

()

¼ a

Pðr

r

8i 6¼ j

08i ¼ j

()

r

8i 6¼ j

08i ¼ j

()

ð6Þ

We see that scaling the pixel intensity by the factor a

merely scales the energy intensity by the same factor a,

whereas adding an offset b is more troublesome, effec-

tively adding a pure energy component corresponding

to an image with constant pixel intensity b. The effect

of the offset and scaling is shown in Fig. 2 with the

channels superimposed. We see that scaling by a factor

of 10 in Fig. 1e has no effect as expected. The original

image in Fig. 1a has a mean value of 77 and a standard

deviation of 47. Figures 1b–d show the effect of pro-

gressively adding offsets of one standard deviation. At

one standard deviation the effect is hardly noticeable

and even at three standard deviations the change is by

no means catastrophic as the channel structure alters

little. We therefore conclude that operational lighting

Physical Analogies for Ear Recognition. Figure 2 Effect of additive and multiplicative brightness changes: (a) original,

(b) 1 std. dev., (c) 2 std. devs., (d) 3 std. devs., (e) scaled 10.

Physical Analogies for Ear Recognition

1085

variation in a controlled biometrics environment will

have little effect. These conclusions are borne out by

the results of the corresponding recognition experi-

ments in Table 1.

Ear Recognition

The technique was validated on a set of 252 ear i mages

taken from 63 subjects selected from the XM2VTS

face database [9] by multiplicative template matching

of ternary threshold convergence maps where levels

less than 1 standard deviation are mapped to 1,

while those greater than one standard deviation map

to +1, and those remaining map to 0. A threshold level

of one standard deviation was chosen experimentally

resulting in the template channel thickness as shown

in Fig. 3c. This ﬁgure also shows a rectangular exclu-

sion zone centered on the convergence magnitude cen-

troid; the centroid of the convergence tends to be stable

with respect to the ear features and thi s approach has

the added advantage of removing unwanted outliers

such as bright spots caused by spectacles. The size of

the rectangle was chosen as 71  51 pixels by adjusting

its proportions to give a good ﬁt for the majority of the

convergence maps. Note how for image 000–2, which

is slightly lower than the other three, the centroid-

centered rectangle has correctly tracked the template

downward.

The inherent automatic extraction advantage was

demonstrated by deliberately extracting or registering

the ears inaccurately, in the sense that the database

consists of 141  101 pixel images, where the ears

have only an average size of 111  73 and are only

roughly located by the eye in the center of these images.

This can be seen clearly in Fig. 3a where we see a

marked variation both in vertical and horizontal ear-

location, and also that there is a generous margin

surrounding the ears. The force ﬁeld technique gives

a correct classiﬁcation rate (CCR) of 99.2% on this set,

whereas running PCA [11] on the same set gives a

result of only 62.4% but when the ears are accurately

extracted by cropping to the average ear size of 111 73,

running PCA then gives a result of 98.4%, thus demon-

strating the inherent extraction advantage. The ﬁrst

image of the four samples from each of the 63 subjects

was used in forming the PCA covariance matrix. Figure 4

shows the ﬁrst four eigenvectors for the 111  73-pixel

Physical Analogies for Ear Recognition. Table 1 Comparison of force field (FFE) and PCA recognition results

Image type Method Passes

Noise

20log

S/N

Correct

classification

rate (%)

Brightness change

by addition

(std dev.) Decidability

141101 with deliberately

poor extraction and

registration

FFE 250/252 Nil 99.2 0 3.432

FFE 251/252 18 dB 99.6 0 3.488

FFE 249/252 12 dB 98.8 0 3.089

FFE 241/252 6 dB 95.6 0 1.886

FFE 250/252 Nil 99.2 1 3.384

FFE 247/252 Nil 98.0 2 3.137

FFE 245/252 Nil 97.2 3 2.846

PCA 118/189 Nil 62.4 0 1.945

111  73 with accurate

extraction and registration

PCA 186/189 Nil 98.4 0 3.774

PCA 186/189 18 dB 98.4 0 3.743

PCA 186/189 12 dB 98.4 0 3.685

PCA 177/189 6dB 93.6 0 3.606

PCA 130/189 Nil 68.8 1 1.694

PCA 120/189 Nil 63.6 2 0.878

PCA 118/189 Nil 62.4 3 0.476

PCA 181/189 Nil 95.6 1 normalized 3.171

PCA 172/189 Nil 91.0 2 normalized 1.91

PCA 166/189 Nil 82.5 3 normalized 1.14

1086

Physical Analogies for Ear Recognition

images. The effect of br ightness chan ge by addition

was also tested where we see that in the worst case

where every odd image is subject to an addition of

three standard deviations the force ﬁ eld results only

change by 2%, whereas those for PCA under the same

conditions fall by 36% , or by 16% for normalized

intensit y PCA, th us conﬁrming that the technique is

robust u nder variable lighting conditions.

These results are presented in Table 1 where Daug-

man’s decidability index [12] combines the mean and

standard deviation of the intra-class and inter-class mea-

surement distributions giving a good single indication of

thenatureoftheresults.Thisindexd

measures, how

well separated the distributions are since recognition

errors are caused by their overlap. The measure aims

to give the highest scores to distributions with the

widest separation between means, and smallest standard

deviations. If the two means are m

and m

and the two

standard deviations are s

and s

then d

is deﬁned as

 m

ﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ

ðs

þ s

Þ=2

: ð7Þ

Note that the best case index for PCA is slightly higher

than the value of 3.43 obtained for the 141  101 images

but this could be attributed to the reduction in data set

size from 252 to 189 and also to the fact that the images

have been extracted for PCA. N oise performance ﬁgures

have also been included where noise has been modeled as

additive noise with a zero mean Gaussian distribution.

The signal to noise ratios of 6 dB, 12 dB, and 18 dB used

are calculated as 20log

(S/N). We see that the technique

enjoys excellent noise tolerance where even for an

extreme noise ratio of 6 dB the performance only

falls by about 3.6%. Interestingly, at a ratio of 18 dB

Physical Analogies for Ear Recognition. Figure 3 Feature extraction for subject 000, row: (a): 141  101 ear images, row

(b): Convergence ﬁelds, row (c): Thresholded convergence maps.

Physical Analogies for Ear Recognition

1087

the recognition rate actually improves over the noise-

less recognition rate, but this must be put down to the

combination of small changes and the random nature

of the noise process. For reference, the corresponding

noise results for PCA under the same conditions have

also been included, where we see that PCA also per-

forms well under noisy conditions but not quite as well

as FFE at 6 dB where the fall is about 4.8%.

Summary

In the context o f ear biometrics, a linear transform

has been developed that transforms an ear image,

with very powerful smoothing and w ithout loss of

information, into a smooth dome-shaped surface

whose special shape facilitates a novel form of feature

extraction that extracts the essential ear signature

without the need for explicit ear extraction. It has

been shown that the technique is robust under vari-

able lighting conditions both by an alysis and experi-

ment. Convergence feature extraction has been

described and it has be en shown that it is a powerful

extension to ﬁeld line featu re extrac tion. The tec h-

nique has been validated by experiment where it has

been shown that it compares favorably with PCA,

especially under variable lighting conditions. In the

process, a contribution has been made to the mount-

ing evidence in suppor t of the recognition potential

of the human ear for biometrics.

Related Entries

▶ Ear Biometrics

▶ Earprints, Forensic Evidence of

▶ Holistic Ear Biometrics

▶ Overview of Ear Biometrics

References

1. Hurley, D.J., Nixon, M.S., Carter, J.N.: Force ﬁeld energy func-

tionals for image feature extraction. In: Proceedings of the 10th

British Machine Vision Conference BMVC99, pp. 604–613. Not-

tingham, United Kingdom (1999)

2. Hurley, D.J., Nixon, M.S., Carter, J.N.: Force ﬁeld energy func-

tionals for image feature extraction. Image Vis. Comput. 20,

311–317 (2002)

3. Hurley, D.J., Nixon, M.S., Carter, J.N.: Force ﬁeld feature extrac-

tionforearbiometrics.Comput. Vis. Image Underst. (CVIU) 98(3),

491–512 (2005)

4. Hurley, D.J., Arbab-Zavar, B., Nixon, M.S.: Handbook of

Biometrics, pp. 131–150. Springer, New York (2008)

5. Burge, M., Burger, W.: Ear biometrics in computer vision. In:

Proceedings of the ICPR 2000, pp. 822–826. Barcelona, Spain

(2002)

6. Yan, P., Bowyer, K.W.: Biometric recognition using three-dimen-

sional ear shape. IEEE Trans. Pattern Anal. Mach. Intell. 29(8),

1297–1308 (2007)

7. Moreno, B., Sanchez, A., Velez, J.F.: On the use of outer

ear images for personal identiﬁcation in security applications.

In: Proceedings of the 33rd IEEE Annual International Carnahan

Conference on Security Technology, 5–7 Oct 1999

8. Chen, H., Bhanu, B.: ‘‘3D free-form object recognition in

range images using local surface patches’’. Pattern Recognit.

Lett. 28(10), 1252–1262 (2007)

9. Messer, K., Matas, J., Kittler, J., Luettin, J., Maitre, G.:

XM2VTSDB: the extended M2VTS database. In: Proceedings

of the AVBPA’99, Washington, DC (1999)

10. Sadiku, M.N.O.: Elements of Electromagnetics, 2nd edn. Saun-

ders College Publishing, Philadelphia, PA, (1989)

11. Turk, M., Pentland, A.: Eigenfaces for recognition. J. Cogn.

Neurosci. (March), 3, 71–86 (1991)

12. Daugman, J.: Biometric decision landscapes. Technical Report

TR482, University of Cambridge Computer Laboratory,

Cambridge, UK (1999)

Physical Analogies for Ear Recognition. Figure 4 First 4 eigenvectors for 111  73 pixel images.

1088

Physical Analogies for Ear Recognition

Physical and Logical Access Control

convergence

The ability to authenticate and authorize a user across

both physical and logical access control systems. The

authentication may mean that a common credential,

such as a biometric, is used across both systems, with-

out the need for a user to re enroll. It may also me an

that the user’s authentication credentials are collective-

ly stored in a mutuall y accessible repository that stores

several biometric and nonbiometric types of authenti-

cation. The harmonization of the authorization of a

user across physical and logical access control systems

implies sing le enterprise policies, which are invoked in

both systems.

▶ Access Control, Physical

Physics-Based Biometrics

▶ Biometric Sample Synthesis

Physics-Based Models

Physics-based models create mathematical equations

that are derived from basic physical principles that can

recreate biometric samples. An example of a physics-

based model is a physics-based vocal tract model.

▶ Biometric Sample Synthesis

Piezoelectric

Piezoelectric is the property of crystals, such as quartz,

which allows them to create electrical voltage when

deformed under the application of a mechanical force.

▶ Digitizing Table t

PIN Replacement

The method of using a biometric to unlock a smart

card or SIM card instead of having to enter a PIN. The

only secure method of PIN replacement is one that can

perform the biometric authentication within the se-

cure element itself, so that it remains in a locked state

until user veriﬁcation is complete.

▶ Transportable Asset Protection

Pitch

The pitch denotes the frequency at which the vocal

cords open and close the larynx to produce the voiced

sounds. The sound then produced resonates according

to the shapes of the different cavities of the vocal appa-

ratus. The pi tch and resonances are generally measured

through a frequency based analysis of the speech signal

to extract information dependent to the speaker.

▶ Transportable Asset Protection

Pixel

Pixel is a single picture element of the image. It never

corresponds to the physical pixel on the camera sensor.

The size of the pixel depends on the image (and the

camera) aspect ratio (the ratio between the image

width and length).

▶ Biometric Sensor and Device, Overview

▶ Image Feature Extraction

▶ Image Formation

Platen

In a ﬁngerprint device, the term platen refers to the

surface of a sensor on which the ﬁnger should roll or

Platen

1089