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

Подождите немного. Документ загружается.

▶ camera point of view. This is really signiﬁcant for

some data to be stated in the record to be coded.

Record Format

After deﬁning the set of requirements for image acqui-

sition, the Standard deﬁnes the way such information

has to be coded and stored within a CBEFF-compliant

wrapper. The structure to be followed is:



A ﬁxed-length (15-byte) general record header con-

taining information aboutthe overall record, with

the following ﬁelds:

– Format identiﬁer (4 bytes with the hexadecimal

value 0x484E4400) and version number (coded

in another 4 bytes)

– Record length (in bytes) including all hand

views within this record (coded in 4 bytes)

– Number of hand views (HGVRs) (1 byte)

– 2 bytes reserved for future use



One or more variable-length ▶ Hand Geometry

View Records (HGVRs), eachcontaining a single

hand silhouette, consisting of:

– A ﬁxed-length (25-byte) hand view header con-

tains the following information:



Length of the HGVR (in 2 bytes)



HGVR index (in 1 byte)



Hand identiﬁer (1 byte), which indicates the

ﬁngers that the system attempts to acquire

within the silhouette, and the view of the

hand (top view of the palm, top view of

the back of the hand, side view from the

thumb side, or side view from the little ﬁnger

side)



Hand Integrity (1 byte), which shows the

identiﬁed problems in the sample acquired

(e.g., ﬁnger missing, misplacement, etc.)



Data resolution in pixels per centimetre (1

byte)



Geometric distortion of the system, as a

signed value incrementing 0.1% (1 byte)



Silhouette quality (3 bytes), with being 0 the

lowest quality and 100 the highest possible

quality, always coded in the lower byte,

while the higher 2 bytes are reserved for

the future use



Camera position relative to the global origin

(1 byte for X position, 1 byte for Y position,

and 1 byte for Z position)



Target position relative to the global origin

(1 byte for X position, 1 byte for Y position,

and 1 byte for Z position)



Silhouette starting point relative to the view

origin (1 byte for X position and 1 byte for Y

position)



Data compression algorithm (1 byte), which

currently refers to only 2 coding methods,

such as 8-way FCC and 4-way FCC

Hand Data Interchange Format, Standardization. Figure 2 Coordinate System linking top and side views, referred to

the 4-FCC directions. Image taken from [1].

670

Hand Data Interchange Format, Standardization



Hand scanning technology (1 byte), giving

information whether the image was ac-

quired using an optical camera, a linear

scanning array, or no information is

speciﬁed



Extended data length (2 bytes)



3 bytes reserved for future use

– Silhouette data, encoded using a Freeman

Chain Code (FCC), either using 8-way FCC,

or 4-way FCC (depending on what is declared

in the ‘‘Data compression algorithm’’ ﬁeld at

the HGVR header

– Extended data (optional), for any application-

speciﬁc or proprietary data used by the system

vendor

For further details refer to the current version of

this International Standard [1]. Curr ent version also

provides a record sample, as well as an informative

annex related to the best practices in this biometric

modality, including hand placement and platen and

optical design.

Other Related Standards

There are other standards related to this technology,

born under ANSI/INCITS scope. This is ANSI/INCITS

396-2005 – ‘‘Information Technology - Hand Geome-

try Interchange Format.’’ This standard is extremely

similar to ISO/IEC 19794-10, where the major techni-

cal differences are:



Within the General Header:

– ANSI/INCITS 396 includes a CBEFF Product

Identiﬁer

– ANSI/INCITS 396 version number is a binary

byte, while in 19794-10 is a 4-byte string



Regarding the View Header:

– ANSI/INCITS 396 has a creation date that was

dropped by ISO 19794-10

– 19794-10 adds a view index that associates

multiple views of the same hand (such as a

top-vie w and side-view captured at the same

time)

– 19794-10 adds a Hand Integrity ﬁeld that indi-

cates which ﬁngers are ok and which are missing/

mangled

– 19794-10 adds a starting-point location linking

the absolute position of the silhouette to the

camera’s optical axis

– 19794-10 supports 4-way or 8-way FCCs, where

ANSI/INCITS 396 only supports 8-way

Due to the fact that the International Standard

ISO/IEC 19794-10 is already available, INCITS has

withdrawn ANSI/INCITS 396-2005.

Summary

To provide interoperability in storing and transmit-

ting hand-geometr y-related biometric information,

one international standard has been developed. Be-

yond this International Standard, other standards

deal w ith conformance and quality control, as well

as interfaces or performance evaluation and reporting

(see relevant entries in this Encyclopedia for further

information).

Related Entries

▶ Biometric Data Interchange Format

▶ Common Biometric Exchange Framework Formats

▶ Hand Geometry

▶ Hand-Geometry Device

▶ Palm Vein

References

1. ISO/IEC: 19794-10:2007 - information technology - biometric

data interchange formats - par t 10: Hand geometry silhouette

data (2007)

2. ISO/IEC: 19785-1:2005 - information technology - common

biometric exchange formats framework - part 1: Data element

speciﬁcation (2005)

Hand Databases and Evaluation

GUANGMING LU

School of Computer Science and Technology,

Shenzhen Graduate School, Harbin Institute of

Technology, Shenzhen, China

Synonyms

Hand geometry; Hand vein; Palmprint database; Palm

vein

Hand Databases and Evaluation

671

Introduction

The human hand is the source of a number of unique

physiological characteristics. The main technologies

for hand recognition fall into three categories:

palmprint techn ologies – those measuring the unique

pattern of prints from the palm of the hand – similar to

a ﬁngerprint; Hand geometry measurements – those

measuring the shape and size of either all or part of the

human hand or ﬁngers; Hand vein patterns – those

measuring the distinct vascular patterns of the human

hand, including hand dorsum vein and palm vein.

Palmprint: A palmprint is deﬁned as the skin pat-

terns of a palm, composed of the physical character-

istics of the skin patterns such as lines, points, and

texture. Palmprints are covered with the same type of

skin as a ﬁnger, but their surface area is much larger

than a ﬁnger tip [1]. Automatic palmprint identiﬁca-

tion systems can be class iﬁed into two categories: on-

line and ofﬂine. An online system captures palmprint

images using a palmprint capture sensor that is directly

connected to a computer for real-time processing. An

ofﬂine palmprint identiﬁcation system usually pro-

cesses previously captured palmprint images, which

are often obtained from inked palmprints that have

been digitized by a digital scanner. Several ofﬂine

palmprint databases were set up for algorithms design

during the early stages of palmprint research and they

yielded promising results. Today most researchers

focus on the online palmprint recognition techniques

and system development. Some public online palm-

print databases are readily available for the researcher

to design efﬁcient recognition algorithms.

Hand geometry: The hand geometry biometric

involves several features including the length, width,

thickness, and surface area of the hand or ﬁngers of a

user [2]. Hand geometry has several advantages over

other biometrics, including small feature size, and

smaller computation requirement as a result of using

low resolution images. In spite of its current wide-

spread use (4.7% market share in 2007), the hand

geometry system suffers from high cost and low accu-

racy. In addition, uniqueness of the hand features is

not guaranteed, making it unsuitable for use in one-to-

many identiﬁcation applications [1].

Hand dorsum vein: The idea of verifying user

identity on the basis of the pattern of subcutaneous

veins on the back of the hand was ﬁrst proposed by

MacGregor and Welford in 1991 [3]. A palm dorsum

vein pattern authentication system named VP-II is

provided by the company I-OnAsia. The system uses

an infrared sensor to capture the thermal image of the

hand dorsum vein pattern. Since no contact with the

device is required, it ensures excellent convenience,

sanitary use, and prevents copying of system-residual

biometrics.

Palm vein patterns: The Fujitsu Laboratories Lim-

ited developed a new type of biometric authentication

technolog y which veriﬁes a person’s identity on the

basis of the pattern of veins in his/her palms [4]. It

makes use of the characteristic of light absorption

difference for oxygenated and deoxidized hemoglobin

in the blood to capture the pattern of the palm vein.

The image of the palm vein would appear as a black

pattern if the palm is illuminated with a near-infrared

light. This is because some of the infrared light is

absorbed by the vein and thus weakens the reﬂected

light back. The company claims that, except for their

size, palm vein patterns are personally unique, and do

not vary over the course of a person’s lifetime.

Palmprint Databases and Evaluation

The earliest paper on palmprint identiﬁ cation [5]

reported in detail the construction of an ofﬂine palm-

print database, which included palmprints of 20 indi-

viduals – 10 prints from both left and right palms

yielding a total of 400 palmprint images. The authors

used 100 dpi resolution to digitize these inked palm-

prints paper sheets with 400  400 pixels to reduce the

computation burden. A sub database, which included

60 special palmprint images, was employed to ascertain

the performance of palmprint line matc hing algorithm

using datum points. This sub database consisted of

20 rotated palmprint images, 35 partial palmprint

images, and 5 palmprint images in which the life line

and head line did not touch each other. Some typical

palmprint images are shown in Fig. 1 and the results of

the datum point detection from special palmprint

images are given in Table 1.

Another ofﬂine palmprint database was developed

and reported by You, Li, and Zhang [6]. It was collec-

ted from 100 individuals w ith two prints from the right

palm, and compris ed a total of 200 palmprint images.

The palmprint images were of 232  232 pixels, with

the resolution of 125 dpi and 256 gray levels. The

palmprint samples were collected from both female

672

Hand Databases and Evaluation

and male adults in the age group of 18–56 years.

Samples of such palmprint images are shown in

Fig. 2. They proposed hierarchical palmprint identiﬁ-

cation via multiple feature extraction. First, a texture

based dynamic selection scheme is proposed to facili-

tate rapid search for the best matching of the sample in

the database. The global texture energy is used to guide

the dynamic selection of a small set of similar candidates

from the database at a coarse level for further proces-

sing. An interesting point based image matching is

performed on the selected similar patterns at a ﬁne

level for the ﬁnal conﬁrmation. The experimental results

from this database show that the average accuracy rate is

95%. Since the majority of samples was ﬁltered out by

the coarse level classiﬁcation, the execution speed of the

ﬁne level feature matching increased signiﬁcantly.

At present many different online palmprint data-

bases have been used in research papers. Among these

[7, 8, 9], the following two public databases are more

representative: The Hong Kong Polytechnic University

(PolyU) Palmprint Database 1.0 and 2.0 [9]. The

PolyU Palmprint Database 1.0 is a subset of 2.0.

The PolyU Palmprint Database 1.0 contains 600

palmprints which were collected from 100 different

palms. Six samples from each of these palms were

collected in two sessions, in other words, three samples

were captured in each session. The palmprints were

of 384  284 pixels and with 256 grayscales. Some-

typical palmprint images in this database are shown

in Fig . 3. Many research papers have been published

that frequently employ this database. The performance

achieved in the literature is ver y high and comparable

with any other established biometric modality. For

example, Wangmeng Zuo et al. [ 10] achieved a recog-

nition rate of 97.67% by using 2DPCA-AMD.

The PolyU Palmprint Database 2.0 contains 7,752

grayscale images corresponding to 386 different palms,

and all the images are available in bitmap format.

Around 20 samples from each of these palms were

collected in two sessions, that is around 10 samples

were captured in the ﬁrst session and the second ses-

sion, respectively. The average interval between the

ﬁrst and the second collection was 2 months. In this

dataset, there were 131 males, and the age distribution

of the subjects was as follows: younger than 30 years

about 86%, older than 50 about 3%, and around 11%

between 30 and 50 years. Numerous papers in

the literature have employed the PolyU Palmprint

Database 2.0. The typical algorith ms that revealed

most promising results are Competitive Coding [11],

RLOC (Robust Line Orientation Coding) [12],

and DoG (Derivative of Gaussian Coding) [13].

The testing results of the three methods based on

the PolyU Palmprint Dat abase 2.0 are given in Fig. 4.

Hand Geometry Databases and

Evaluation

The hand ge ometry features can be typically extrac-

ted from a binary image. Therefore, low-resolution

Hand Databases and Evaluation. Figure 1 Examples of datum points determination: (a) normal palmprint, (b) rotated

palmprint, (c) incomplete palmprint, and (d) the life line and head line unintersection [5].

Hand Databases and Evaluation. Table 1 Experimental

results of datum point determination in special palmprint

images

Classification R I U

Experiment images 20 35 5

Accurate determination images 19 33 4

Rate of accuracy (%) 95 94 80

R rotated image; I incomplete image; U life line and head line

unintersection

Hand Databases and Evaluation

673

Hand Databases and Evaluation. Figure 2 Samples of different palmprint patterns with distinctive texture features

(a) strong principle lines, (b) full of wrinkles, (c) less wrinkles, and (d) strong wrinkles [6].

Hand Databases and Evaluation. Figure 3 Some typical palmprint images from PolyU palmprint database 1.0.

Hand Databases and Evaluation. Figure 4 The ROC curves of the three typical identification methods based on the

PolyU palmprint database 2.0.

674

Hand Databases and Evaluation

imaging is employed to gain the advantage associated

with faster processing. The hand geometry database

introduced by Savic and Pavesi [14], includes the gray-

scale hand images of 100 individuals (68 males and

32 females) with 10 images of the right hand and

10 images of the left hand (a total of 2,000 images).

The average age of the subjects was 36 years; the oldest

was 78 years and the youngest was 21 years, all subjects

belonged to the same ethnic group. The images were

collected in two separate sessions over a period of

3 months. As the left and right hands are different,

the left hand images were mirrored and used as the

right hand images of ‘‘new’’ people. In this way, 200

image classes with 10 images per class were obtained.

The geometrical parameters extracted from the hand

are illustrated in Fig. 5(a), and the experimental result

is presented in Fig. 5(b).

Kumar et al. [ 15] developed a hand database

captured by a digital camera, in which they collec-

ted 1,000 images of the left hand from 100 subjects,

i.e., 10 samples from each subject. The geometry

features are deﬁned in Fig. 6(a). Based on this

method, the testing results are presented in Fig. 6(b)

when they used the ﬁrst ﬁve images from each

subject for training and the remaining for testing.

One of the earlie st hand geometry database con-

struction was repor ted by Sanches-Reillo et al. [16],

which comprises 200 hand images obtained from

20 people of different ages, sex, profession, and living

style. Ten samples were collected from each person on

different days over a period of 3 months. Experimental

results show that the proposed method can achieve up

to 97% accuracy in classiﬁcation by using Gaussian

Mixture Modeling.

Hand Dorsum Vein Databases and

Evaluation

Hand dorsum vein pattern has been widely researched

during the last decade. Some typical hand dorsum vein

images obtained from low-cost near infrared imaging

are shown in Fig. 7. The researchers designed their own

capture devices to collect hand dorsum vein images. So

far, there is no public hand dorsum vein database

available to researchers. Here, the attempt is to intro-

duce some typical hand do rsum databases from pub-

lished research papers.

Lin and Fan [17] obtained 960 palm dorsa thermal

images from 32 volunteers. The volunteers in this set

of veriﬁcation experiments included 29 male adults

and 3 females. Each of the thermal images of

640  480 pixels was captured under varying light

conditions (even in the dark). The authors obtained

an equal error rate of 2.3%. Another hand dorsum vein

database which includes 300 images from 100 persons

Hand Databases and Evaluation. Figure 5 (a) Geometrical parameters of the hand, (b) the EER as a function of the

regularization parameter [14].

Hand Databases and Evaluation

675

of different ages and gender has been employed by

Kumar and Prathyusha [18]. This database was devel-

oped by using a low-cost near infrared camera, tradi-

tionally utilized for surveillance, using contactless

image acquisition. The volunteers were requested to

present their folded right hand (palm dorsal surface)

near the imaging window such that the knuckle tip of

the middl e ﬁnger remained at the top. The authors

[18] obtained an equal error rate of 1.14% by integrat-

ing knuckle shape information into this contactless

image database. Sang-Kyun Im et al. [19] developed a

hand dorsum vein database comprising 5,000 grayscale

images. Each image was of 160  120 pixels. They

evaluated the database by using CSD codes and

achieved a reliability accuracy of 99.45%. By using

the FPGA (ﬁeld programmable gate array) device, the

whole veriﬁcation process required only 150 ms per

person.

Palm Vein Database and Evaluation

Research on palm vein technologies is fairly recent and

employs low-cost near infrared imaging to obtain palm

vein images. The Fujitsu Laboratories Limited developed

a palm vein system by testing 1,400 palm proﬁles of

Hand Databases and Evaluation. Figure 7 Typical hand dorsum vein images.

Hand Databases and Evaluation. Figure 6 (a) Deﬁned geometrical features, (b) the ROC of the testing results (note the

hand geometry curve) [15].

676

Hand Databases and Evaluation

700 individuals, the system achieved a false rejection rate

of 1% and a false acceptance rate of 0.5% with an equal

error rate of 0.8% [4]. The palm vein image database

used by Wang et al. [20] was constructed from 120

subjects and each image was of 768  576 pixels. This

database also contains palmprint images as the

employed image acquisition device can simultaneously

acquire palm vein images from near infrared imaging

and palmprint images from visible illumination.

Summary

Hand based biometric technologies have attracted a lot

of attention both in research and industry. An increas-

ing number of hand based biometric systems have

appeared in the literature. There is a lot of interest in

research on hand based biometrics for the simulta-

neous acquisition of several hand based traits. These

multimodal biometric technologies try to combine all

the features (palmprint, hand geometry, hand vein,

palm vein) of the hand to build a higher accuracy

biometric system. Such multimodal systems offer

very high resistance against spoof attacks and have

diverse applications.

Related Entries

▶ Hand Geometry

▶ Hand Vein

▶ Palm Vein

▶ Palmprint

References

1. Zhang, D.: Palmprint Authentication, Kluwer, Dordrecht (2004)

2. Kumar, A., Zhang, D.: Hand geometry recognition using

entropy-based discretization. IEEE Trans. Inf. Forensics and

Secur. 2(2), 181–187 (2007)

3. MacGregor, P., Welford, R., Veincheck, R.: Imaging for security

and personnel identiﬁcation. Adv. Imaging. 6(7), 52–56 (1991)

4. Fujitsu Laboratories Limited. Contactless Palm Vein Pattern

Biometric Authentication System. http://pr.fujitsu.com/en/

news/2003/03/31.html

5. Zhang, D., Shu, W.: Two novel characteristics in palmprint

veriﬁcation: datum point invariance and line feature matching.

Pattern Recogn. 32(4), 691–702 (1999)

6. You, J., Li, W., Zhang, D.: Hierarchical palmprint identiﬁcation

via multiple feature extraction. Pattern Recogn. 35(4), 847–859

(2002)

7. http: //vis graph.cs.ust.hk/biometrics/Visgraph_web/index.html.

Last Accessed 27 March, 2009

8. http://www.cbsr.ia.ac.cn/china/Palmprint%20Databases%20CH.

asp. Last Accessed 27 March, 2009

9. http://www.comp.polyu.edu.hk/~biometrics. Last Accessed 27

March, 2009

10. Zuo, W.M., Zhang, D., Wang, K.Q.: Bidirectional PCA with

assembled matrix distance metric for image recognition. IEEE

T. Syst. Man Cy. B. 36(4), 863–871 (2006)

11. Kong, A.W.K., Zhang, D.: Competitive coding scheme for palm-

print veriﬁcation. Pattern Recogn. 1 , 520–523 (2004)

12. Jia, W., Huang, D., Zhang, D.: Palmprint veriﬁcation based

on robust line orientation code. Pattern Recogn. 41(5),

1054–1513 (2008)

13. Wu, X., Wang, K., Zhang, D.: Palmprint texture analysis

using derivative of gaussian ﬁlters. In: 2006 International Confer-

ence on Computational Intelligence and Security, Berlin, 751–754

(2006)

14. Savic, T., Pavesi, N.: Personal recognition based on an image of

the palmar surface of the hand. Pattern Recogn. 40, 3152–3163

(2007)

15. Kumar, A., Wong, D.C.M., Shen, H.C., Jain, A.K.: Personal

authentication using hand images. Pattern Recogn. Lett. 27,

1478–1486 (2006)

16. Sanches-Reillo, R., Sanchez-Avila, C., Gonzalez-Marcos, A.: Bio-

metric identiﬁcation through hand geometry measurements.

IEEE T. Pattern Anal. 22(10), 1168–1171 (2000)

17. Chih-Lung, L., Kuo-Chin, F.: Biometric veriﬁcation using

thermal images of palm-dorsa vein patterns. IEEE T. Circ. Syst.

Vid. 14(2), 199–213 (2004)

18. Kumar, A., Prathyusha, V.: Personal authentication using hand

vein triangulation. In Proceedings of the SPIE, vol. 6944,

pp. 69440E–69440E–13 (2008)

19. Sang-Kyun, I., Hyung-Man, P., Young-Woo, K., Sang-Chan, H.,

Soo-Won, K., Chul-Hee, K.: A biometric identiﬁcation system

by extracting hand vein patterns. J. Korean Phys. Soc. 38(3),

268–272 (2001)

20. Wang, J.-G., Yau, W.-Y., Suwandy, A., Sung, E.: Person recogni-

tion by fusing palmprint and palm vein images based on lapla-

cianpalm representation. Pattern Recogn. 41(5), 1514–1527

(2008)

Hand Geometry

RAUL SANCHEZ-REILLO

University Carlos III of Madrid, Avda. Universidad,

Madrid, Spain

Synonyms

Hand biometrics; Hand shape biometrics

Hand Geometry

677

Definition

Biometric modality based on identifying a person by

the shape of his or her hand. In its basic form, it is

based on taking a photograph of the user’s hand while

placed on a surface, and after a

▶ contour detection,

ﬁnding singular points and taking measurements

among them.

Introduction

Hand Geometry is considered as a medium-proﬁle

biometric modality which reaches a really high level

of user acceptance with low computational cost. Not

being one of the ﬁrst biometric modalities, it has

gained great popularity due to the success of some

commercial products, at the end of the twentieth cen-

tury. In fact, the commercial product from Schlage

Recognition Systems, known as HandKey II [1], was

one of the most sold at the beginning of the 2000s,

especially for physical access control systems and time

and attendance control.

As mentioned belo w, after some initial works, other

scientists have continued researching on other algorithms

and more comfortable means of using this technology.

Error rates achieved are not as low as those modalities

considered as high-performance ones (e.g., ﬁngerprint,

iris or vascular). In order to gain applicatibility , some

researchers have included this technology in multimodal

biometric systems, reducing error rates, and gaining in

▶ usability and user acceptance.

Basics and Initial Works

Hand Geometry biometrics is based on the measure-

ment of the shape of the contour of the hand [2],

including ﬁnger widths at several points, ﬁnger lengths,

palm shape, deviation angles, etc. Main idea comes

from the Bertillon system (http://en.wikipedia.org/

wiki/Bertillon) used during the late nineteenth century

to identify prisoners. But it was not till 1997 that the

ﬁrst paper in a scientiﬁc journal is found. In such

paper, among may other interesting things, Golfarelli

et al. [3] outline a system based on a semi-opaque

plastic material with some ﬁxed

▶ pegs to guide posi-

tioning of the hand. With a CCD camera located over

the hand, and some light located under the surface

plate, a high contrast image of the user’s hand is

obtained. As to acquire also the lateral projection

of the hand , the system is replicated on the side, and

a 45 mirror is placed to project such image to the

same camera. The counterlight image allows a very

easy contour detection of the hand, and from there

17 geometrical features are extracted. Figure 2 illus-

trates image acquisition and feature extraction steps

from this work.

In 1999 and in 2000 two pape rs were published

detailing this biometric modality. They were written

Hand Geometry. Figure 1 HandKey II device and illustration of its use as a door lock in a physical access control system

[1] (Images published under authorization of Schlage Recognition Systems).

678

Hand Geometry

by Jain et al. [4] and Sanchez-Reillo et al. [5]. In this

last work, the device developed is also based on a CCD

camera located over the hand of the user. But differ-

ently from the Golfarelli approach [3], here the hand is

located over an opaque peg-oriented surface painted in

blue (see Fig. 3). The reason for the platform to be blue

is that the human skin, no matter the race, has a ver y

low portion of blue component. Therefore discarding

all blue component in the RGB, allows an easy way to

eliminate all background information. As in Golfarelli

et al. system [3], also a mirror is placed to obtain the

lateral view of the hand. In contrast, the illumination

demands of this new system were lower, as only the one

coming from the camera built-in ﬂash was employed.

From the image sample acquired, as mentioned

above, the background surface is removed by eliminat-

ing the blue component of the image. In fact, in order

to obtain a better output for next stage, the back-

ground removal is done by the following formulae

(cropping all negative values to 0):

Hand Geometry. Figure 2 Illustration of the Hand Recognition system designed by Golfarelly et al., including a sample

of the photographs taken and the geometrical measurements extracted. Images taken from [3]. ßIEEE.

Hand Geometry. Figure 3 Prototype developed in [5] and measurements taken: (a) General view of the prototype;

(b) Positioning of the hand; (c) Sample taken; (d) Geometric features. Images extracted from [5]. ßIEEE.

Hand Geometry

679