Palmprint and Handgeometry Recognition using FAST features and Region properties

International Journal of Science and Engineering Applications
Volume 3 Issue 4, 2014, ISSN-2319-7560 (Online)
www.ijsea.com 75
Palmprint and Handgeometry Recognition using FAST features
and Region properties
Swapnali G. Garud1
K.V. Kale2
Department of CS and IT, Department of CS and IT,
Dr. Babasaheb Ambedkar Marathwada Dr. Babasaheb Ambedkar Marathwada
University, Aurangabad, (M.S.), India. University, Aurangabad, (M.S.), India.
Abstract: Biometrics recognition system is more reliable and popular. In this paper we describe a palmprint and handgeometry based person
identification consisting of three main steps - preprocessing techniques such as morphological operations. The feature extraction techniques
such as FAST feature algorithm and region properties is used to independently extract palmprint and handgeometry features. Feature matching
with euclidean distance classifier. These techniques are more reliable and faster than traditional techniques used. We finally conclude that the
proposed methodology has better performance .This is supported by our experiments which are able to achieve recognition rate for palmprint
100 % and for handgeometry 93.75 %.
Keywords: Multimodal, FAST features, Region properties, Euclidean
1.INTRODUCTION
In real life effective access control system is challenging
task. The biometric authentication system becomes very popular
because it uses behavioral and physiological characteristics to
uniquely identify the individual [1]. Biometrics authentication
system useful in various applications such as physical access control,
security, monitoring which is more secured than traditional password
based security systems because it is not proper practice each time to
remember long passwords hence password based authentication
system is referred as the weak authentication mode and the
Biometrics can be employed on various traits like fingerprint,
palmprint, handgeometry, iris, face, voice, signature, etc which are
unique for every individual hence are referred as strong
authentication mode [2]. Biometrics system has two types i.e.
verification systems and identification systems [3]. Many
comparisons are required for this system Biometrics is of three
modes i.e. unimodal biometrics which can identify individual by
using single trait. Second is bimodal biometrics in which
identification is done with fusion of two modalities and the other is
multimodal biometrics which uses combination of multiple traits for
identification purpose of human. A unimodal biometric may fails to
be accurate enough for the identification of a user population and
there is one more possibility of failure if physical characteristics of a
person for the selected biometric is not available. The chance of any
two people having the same characteristic will be optimized by
highly unique features [4]. By combining information from different
biometrics modalities we can achieve higher and more consistent
performance levels [5]. A multimodal biometric system requires an
integration scheme to fuse information obtained from the individual
modalities. Biometrics system performance depends on quality of
image [6]. There are different level of fusion available in
biometrics. Amongst that feature extraction level is widely used
because many observers prove that this level fusion produces better
results. Usually, the performance of the biometric system is given by
the accuracy of the system
In this paper new method is provided for personal
authentication using palmprint and handgeometry that are
simultaneously acquired from a single hand image. The database
contains images of subjects for left and right hand . Each of these
palmprint images are used to extract specific features. Thus the
palmprint and hand geometry features of an individual are obtained
from the same hand image.
The palm is the inner surface of a hand between the wrist
and the fingers. The palm has unique features and provides a larger
area so the more distinctive features can be generated to improve the
performance of recognition system. There are different features that
exists on a palm such as principle line, wrinkle line, delta point. The
feature of palmprint is quite stable and specific because there is a
little change in a long time. They can only be generated from a high
resolution image, hard to be faked.
Hand Geometry gives the geometric structure of the hand.
Hand geometry is suitable for integration with other biometrics.
H a n d G e o m e t r y i s a b i o m e t r i c
k e y w i t h m e d i u m l e v e l o f
i n d i v i d u a l i z a t i o n . T h e r e a r e
d i f f e r e n t f e a t u r e s t h a t c a n b e
e x t r a c t e d a n d u s e d a s k e y s u c h a s
f i n g e r w i d t h a n d l e n g t h , hand height,
width, palm height, palm width, etc [7]. The feature of hand
geometry is relatively simple and easy to use but it is not invariant
due to a period of time.
The rest of the paper is organized as follows - Section II
gives background of related work in literature. Section III gives
proposed method which contains image acquisition, preprocessing,
feature extraction techniques. Section IV discussed the experiments
and results. Section V summarily devoted to conclusion. Section VI
gives acknowledgement.

www.ijsea.com 76
2.RELATED WORK
Slobodan Ribaric, et al [1] gives a bimodal biometric verification
system for physical access control based on the features of the
palmprint and the face, palm matching is based on the adapted
HYPER method. And for face the K–L transform is used for
matching. bimodal system can achieve an EER (equal error rate) of
3.08% for T=0.748 and the minimum TER (total error rate) = 5.94%
for T = 0.8. Antonia Azzini, et al [2] given idea about using a fuzzy
control system to manage a multi-modal authentication system,
checking the identity of a user during the entire session. The first
biometric acquisition takes matching score 0.725and the second
biometric acquisition takes score 0.4860. Teddy Ko, [3] gives
various scenarios in multimodal biometric systems using fingerprint,
face and iris recognition, the levels of fusion that are possible and the
integration strategies that can be adopted to fuse information and
improve overall system accuracy. How the image quality of traits
will affect the overall identification accuracy and the need of staffing
for the secondary human validation . V. C. Subbarayudu, et al [4]
proposed general working of multimodal biometrics system with Iris
and Palmprint and fusion is done at the matching score level by Sum
Rule technique with recognition rate is 96.6%. Andrew Teoh, et al
[5] introduced k-Nearest Neighbourhood (k-NN) based classifiers are
adopted in the decision fusion module for the face and speech experts
with Recogntion rate is 80.33%. Anil K. Jain, et al [6] described an
automated fingerprint recognition system and listed key challenges
and research opportunities in the field. The recognition rate is 95%.
Fan Yang, et al [7] fingerprint, palm-print and hand-geometry are
combined for person identity verification. Wavelet transform to
extract the features from fingerprint and palm-print is used and hand-
geometry feature (such as width and length) is extracted after the pre-
processing phase. Feature level fusion and match score fusion
together for identity.The weight values are calculated based on total
minimum error. i.e. For weight1- 0.75,weight2- 0.25.e X. Wu., et al
[8] proposed a palm print recognition system by extracting features
using Sobel operators and using Hidden Markov Models (HMM) as
classifiers. Ajay Kumar, et al [9] attempts to improve the
performance of palmprint-based verification system by integrating
hand geometry features. These features are then examined for their
individual and combined performances. The recognition rate is
98.3%.
Harpreet Singh, et al [10] have given iterative fuzzy
approach for obtaining fused images Entropy values are provided in
result as for Fuzzy algorithm entropy is 5.30 and for neuro fuzzy
algorithm 4.89. Chun Wai Lau, et al [11] presents a multi-biometric
verification system that combines speaker verification, fingerprint
verification with face identification and equal error rates (EER) are
4.3%, 5.1% and the range of (5.1% to 11.5%) for matched
conditions in facial image capture respectively. K. Ito, et al [12]
suggested Multi-scale wavelet decomposition of palmprint images
and using mean of each wavelet sub-block has been suggested .
M. Wang, et al [13] proposed 2D PCA and 2D LDA
over conventional PCA have been reported to be better for
palmprint recognition. V. Conti, et al [14] have proposed
multimodal biometric system using two different fingerprints. The
matching module integrates fuzzy logic methods for matching score
fusion. Both decision level fusion and matching score level fusion
were performed.
Kornelije Rabuzin, et al [15] had suggested active rules in fuzzy
logic are used for effective decision making in person identification.
The recognition rate is 97%.
Gawande, et al [16] used log Gabor filter can be used to
extract the feature vectors from both Iris and Fingerprint and then
they are concatenated. The phase data from 1 D log Gabor filters is
extracted and quantized to four levels to encode the unique pattern
of Iris and Fingerprint into bitwise biometric template. Hamming
distance (HD) is used to generate a final match score. Yong Jian
Chin, et al [17] proposed a multimodal biometrics system in which
2D gabor filter is used to extract features. The recognition rate is
98%. Cheng Lu, et al [18] suggested idea which utilizes two or
more individual modalities, like face, ear, and fingerprint, to
improve the recognition accuracy by new dimensionality reduction
method called Dimension Reduce Projection (DRP). The
recognition rate is 95.8%. Nicolas Tsapatsoulis, et al [19] presented
an identification and authentication system based on hand geometry
which used POLYBIO hand database. The recognition rate is 95%.
Anil K. Jain, et al [20] given an overview of biometrics,
emerging biometric technologies and their limitations, and examines
future challenges. Mohammad Imran, et al [21] proposed a new
hybrid approach to verification aspect of a multibiometrics system,
comparative analysis with traditional approaches such as
multialgorithmic and multimodal versions of the same. The average
EER of hybrid approach from different levels of fusion is 3.87%
which shows that hybrid approach yields lower average EER.
Mohamed K. Shahin, et al [22] introduced a multimodal biometric
system (MMBS) based on fusion of whole dorsal hand geometry and
fingerprints that acquires right and left (Rt/Lt) near-infra-red (NIR)
dorsal hand geometry (HG) shape and (Rt/Lt) index and ring
fingerprints (FP). Accuracy rate is 99.71%. S. Palanikumar, et al [23]
presented approach for enhancing palmprint image. The
enhancement is based on curvelet which preserves the fine features
without noise. The result gives high PSNR (Peak Signal-to-Noise
Ratio) value for the Curvelet method. i.e. 38.1047 .
Feifei CUI, et.al. [24] proposed multimodal biometrics
recognition based on score level fusion of fingerprint and finger
vein. Recogntion rate is 98.74%. Romaissaa Mazouni, et al [25]
proposed a comparative study of several advanced artificial
intelligence techniques (e.g. Particle Swarm Optimization, Genetic
Algorithm, Adaptive Neuro Fuzzy Systems, etc.) as to fuse
matching scores in a multimodal biometric system is provided. The
fusion was performed under three data conditions: clean, varied and
degraded. Some normalization techniques are also performed prior
fusion so to enhance verification performance. The population based
techniques (PSO, GA) gave very good results. Nishant Singh, et al
[26] presents an efficient multimodal biometric system based on 4
slap fingerprint images. The system utilizes 4 slap fingerprint
scanner to simultaneously collect fingerprints of multiple fingers on
a hand in one image. Decision threshold is 0.9869 and FAR is
5.08%. Ashutosh Kumar, et al [27] suggested the new approach
where the palmprint images are mapped to Eigen-space and a robust
code signature is generated from different camera snapshots of the
same palm to incorporate tonal and lighting variations. To enable
real-time identification, the signature is represented by a low
dimensional feature vector to reduce computational overheads. P.U.
Lahane, et al [28] given the comparison of data base template and
the input data is done with the help of hamming-distance matching
algorithm. If the templates are matched we can allow the person to
access the system. Gabor filter is used for fingerprint.
Krishneswari, et al [29] proposed to investigate the
performance of multimodal biometrics using palm print and
fingerprint. Features are extracted using Discrete Cosine Transform
(DCT) and attributes selected using Information Gain (IG). Results
shows an average improvement of 8.52%. D. Y. Liliana, et al [30]
studied about biometrics of palm for identification system using
block-based line detection for palm print feature extraction process,
and chain code solved the hand geometric feature extraction. We
combined those two respective features and recognized it using
Dynamic Time Warping (DTW) method which was able to measure
the distance between two different features. The accuracy rate is
89%. Gawande, et al [31] gives use of the log Gabor filter to extract

www.ijsea.com 77
the feature vectors from both Iris and Fingerprint and then they are
concatenated. Finally the phase data from 1 D log Gabor filters is
extracted and quantised to four levels to encode the unique pattern of
Iris and Fingerprint into bitwise biometric template. Hamming
distance (HD) is used to generate a final match score. Experimental
results was verified on database of 50 users accounting to FAR = 0%
and FRR = 4.3%. M. Dale, et al [32] presented palm texture using
transform features and hand geometry features are represented as
distances between different boundary points. The final decision is
made by fusion at decision level. S. Rao, et al [33] suggests image
fusion using fuzzy and neuro fuzzy logic approaches utilized to fuse
images from different sensors, for enhancing visualization.
Mohammad Abdolahi, et al [34] proposed two biometric traits such
as iris and fingerprint which uses Decision level fusion and Fuzzy
logic as technic for the effect of each biometric result combination.
Recognition rate is 71%.
There are many challenges in traditional system such as they
requires more time for recognition. This research tried to overcome
this problem by suggesting new algorithm for finding the features of
palm which is very much faster with higher recognition rate.
3.PROPOSED METHOD
The proposed method includes various steps such as image
aquisition from sensor, preprocessing operation to enhance the
quality of image and feature extraction process to identify the
features of an image. Finally, matching is done on the basis of
specific features with database image and decision is made for
identification.
Figure 1. Block diagram of proposed system
The figure 1 shows the block diagram of proposed
biometrics identification system.
3.1 Image Acquisition
We used KVKR Multimodal Biometrics database#
This
database contains images of different modalities such as palm ,
handgeometry, fingerprint, iris, face, voice, signature, HRV, knuckle,
etc. of 150 subjects which belongs to age group of 20 to 30 years old
. Out of which data for 7 subjects is used in this experiment. From this
total 112 images for palmprint and 112 images for handgeometry of
7 subjects.
For extraction of palmprint images whole
handgeometry images are employed. Palmprint images are acquised
from the center of rectangle that can enclose the whole area of
interest in palm. These center coordinates are used to extract a square
palmprint region of fixed size. Out of these total 112 images of
palmprint from database 8 images for left and right hand i.e. 16
images of every subject.
For extraction of handgeometry images we used Laserjet
scanner and they are of JPEG format. The users were only
requested with two preconditions first is their fingers do not touch
each other and second is most of their hand touches the scanner
surface. Handgeometry database is collected. We used total 112
images of handgeometry from database 8 images for left and right
hand. i.e. 16 images for each subject.
3.2 Preprocessing techniques
For palmprint images we used center region of handgeometry
images of specific size. These images are also colored images and
we convert them firstly into grayscale images. Then specific
threshold value is set for images for further processing.
For handgeometry images we used color hand
images and they are firstly converted into grayscale format. Then
we used morphological opening to estimate the background
illumination. The opening operation has the effect of removing
objects that cannot completely contain the structuring element.
Then grayscale image converted into a binary image by using
thresholding. After that background noise is removed. Then we
found all the connected components in the binary image. The size
of the objects has major concern in the accuracy of results.
3.3 Feature extraction techniques
For palmprint we extracted features such as corners
points in grayscale images with the Features from Accelerated
Segment Test (FAST) algorithm to find feature points. From these
corner points interest point descriptors are extracted. These corner
points are common for both images in the form of Index Pair. This
algorithm also gives value of total number of Index pairs in images.
Greater the value of index pair indicates that the images belongs to
same subject and smaller value of index pair shows that the images
belongs to different subject. The descriptors are extracted feature
vectors and their corresponding locations, from a binary or intensity
image. The function derives the descriptors from pixels surrounding
an interest point. These pixels describe and match features specified
by a single-point location. Each single- point specifies the center
location of a neighborhood. The method used for descriptor
extraction depends on the class of the input points such as
SURFpoints, MSERobjects, corner points. FASTfeatures technique
uses parameters as image, name and its scalar threshold value in the
range (0,1). Name is MinContrast . It gives minimum intensity
difference between corner and surrounding region, specified as the
# is explained in Section 8. Annexure.

www.ijsea.com 78
comma-separated pair consisting of MinContrast. The minimum
intensity represents a fraction of the maximum value of the image
class. Increasing the value reduces the number of detected corners.
The default value is 0.2. A standard threshold value used in this
experiment is 0.031 because at this particular threshold value we get
maximum number of matching index pairs in images.
Table 1. Feature matrix palmprint palmprint images for Left
hand
Sub
Index
Pair1
Index
Pair2
Index
Pair3
Index
Pair4
Index
Pair5
Index
Pair6
Index
Pair7
Sub1 31.6 33 31 29.4 24.14 29.6 27.4
Sub2 16.9 31 33.8 26.3 41.12 41.4 31.6
Sub3 18.6 17.37 65 69.6 7 18.8 17.8
Sub4 15.5 15.12 5.12 5.5 11.25 10.8 56.5
Sub6 2.62 4.87 3.25 3.25 3.62 3.25 4
Sub7 1.5 1.25 1.63 1.75 1.12 2 2.37
Table 2. Feature matrix palmprint palmprint images for
Right hand
Sub
Index
Pair1
Index
Pair2
Index
Pair3
Index
Pair4
Index
Pair5
Index
Pair6
Index
Pair7
Sub1 23.38 25.50 90.50 11.50 33.13 28.00 7.88
Sub2 16.88 31.00 33.75 26.25 41.13 41.38 31.63
Sub3 5.13 5.00 8.25 7.50 8.38 9.88 9.63
Sub4 23.63 21.25 14.25 15.75 12.75 15.25 24.25
Sub5 9.88 15.00 19.63 14.75 20.25 16.25 15.00
Sub6 10.88 12.75 14.63 12.63 15.75 13.88 14.50
Sub7 3.25 3.25 2.63 3.25 4.38 1.38 0.63
For handgeometry we extracted features by using the region
properties which measures properties of image regions. In this
experiment we extracted features such as area, boundingbox, height
of hand, perimeter, majoraxislength, minoraxislength by using the
region properties of connected component of binary image of hand.
The connected component gives the number of pixels. The area
feature gives the actual number of pixels in the region. The perimeter
gives distance around the boundary of the region. The Boundingbox
gives smallest rectangle containing the region. It also gives the
height of region. The majoraxislength specifies the length of the
major axis of the ellipse that has the same normalized second central
moments as the region and minoraxislength gives the length of the
minor axis of the ellipse that has the same normalized second central
moments as the region. By all these region properties we can easily
discriminate the features of different images. First 6 samples were
considered as training samples and mean is taken for those samples
for all images for each subject. And remaining 2 samples of all
features of each subject is considered as testing sample. In table 3
and 4 F1– F6 shows feature 1 to feature 6.
Table 3. Feature matrix handgeometry images for Left hand
Sub
Area Perimeter
Bounding
Box
height
Major
Axis
length
Minor
Axis
length
F1 F2 F3 F4 F5 F6
Sub1 737798.83 11950.85 1413.83 1405.67 1341.01 934.16
Sub2 450250.50 7575.32 1137.17 1126.33 1057.40 770.87
Sub3 395181.33 6495.84 1031.83 1031.83 982.47 690.55
Sub4 459746.50 7010.65 1109.33 1109.33 1084.23 758.53
Sub5 409272.83 7206.35 1117.83 1055.33 980.00 774.93
Sub6 447932.83 6273.44 1114.33 1114.33 1062.16 703.51
Sub7 432571.67 8010.98 1020.00 1020.00 1022.54 766.52
Table 4. Feature matrix handgeometry images for Right
hand
Sub
Area Perimeter
Bounding
Box
height
Major
Axis
length
Minor
Axis
length
F1 F2 F3 F4 F5 F6
Sub1 746972.50 12162.86 1420.25 1412.67 1338.32 981.31
Sub2 460402.17 7306.92 1116.75 1114.83 1074.19 782.85
Sub3 402758.67 6349.36 1008.00 1008.00 982.88 728.61
Sub4 445831.17 6822.90 1156.67 1086.67 1038.19 811.78
Sub5 403517.67 6064.52 1015.50 1015.50 952.13 752.21
Sub6 444826.33 6331.65 1109.33 1109.33 1063.97 709.75
Sub7 457211.83 7846.00 1068.67 1068.00 1064.75 788.66
4. EXPERIMENT AND RESULT ANALYSIS
4.1 Palmprint Recognition
In palmprint we used 112 images for 7 subject of both left
and right hand. From 112 we used 98 for training and remaining 14
for testing matrix. We test at least two images at a time for palm. By
comparing results of these two images we can easily recognize the
particular subject. If the image sample belongs to same person then
the both image contains maximum no of matching corner points in
common called as index pair. If the images doesn’t belongs to same
person then they have no matching points in common or negligible
matching points in common. We can compare the test image against
no of images at the same time with this process. The resultant
matrix contains total no of index pairs for each pair of tested image
samples. This test gives appropriate idea about the test sample
belongs to which subject. Afterwards this matrix for further analysis
and ease of use is reduced to the classification matrix which
contains the total no of samples correctly classified in particular
class for each subject. And ‘x’ entry indicates that no match in
corresponding class. The table 14 shows classification matrix for
palmprint samples.

www.ijsea.com 79
Table 5. Classification matrix for palmprint samples
Test
Sample
Sub1 Sub2 Sub3 Sub4 Sub5 Sub6 Sub7
Class-
1
Class-
2
Class-
3
Class-
4
Class-
5
Class-
6
Class-
7
Sub1-8 8 x x x x x x
Sub2-8 x 8 x x x x x
Sub3-8 x x 8 x x x x
Sub4-8 x x x 8 x x x
Sub5-8 x x x x 8 x x
Sub6-8 x x x x x 8 x
Sub7-8 x x x x x x 8
Table 6. Recognition rate for palmprint samples
Sub
Total no
of sample
tested
Correct
classified
Miss
Classified
RR
Sub1 8 8 0
100%
Sub2 8 8 0
Sub3 8 8 0
Sub4 8 8 0
Sub5 8 8 0
Sub6 8 8 0
Sub7 8 8 0
From table 6 it is observed that all samples of palmprint are
classified into corresponding classes correctly. So the
recognition rate achieved for palmprint images is 100 %.
4.2 Handgeometry Recognition
In Handgeometry we used 112 images for 7 subjects of both left
and right hand. From 112 we used mean of 6 samples of each subject
for every feature and remaining 14 images for testing matrix. First we
calculate the feature matrix then the pairwise Euclidean distance
between training matrix and testing matrix is calculated . Mean of
each sample is tested with 7 and 8th
sample of each image. The
resulting distance matrix for left hand as shown in table 7.
Table 7. Distance matrix for handgeometry images of Left
hand
Sub Sub1 Sub2 Sub3 Sub4 Sub5 Sub6 Sub7
S1
7
3237.69 284436.30 339515.14 274949.03 325415.50 286773.03
302109.23
S1
8
17642.45 269940.60 325020.00 260454.83 310919.47 272280.03
287612.29
S2
7
288754.75 1172.74 53907.36 10682.52 39806.82 1721.05
16513.26
S2
8
289191.26 1609.56 53471.03 11120.40 39370.11 1491.84
16075.74
S3
7
339392.12 51818.33 3499.28 61318.70 10856.12 49523.31
34139.46
S3
8
337535.79 49961.16 5259.06 59461.28 8999.66 47665.82
32282.82
S4
7
292207.52 4688.19 50465.13 14187.68 36358.05 3059.86
13046.86
S4
8
276100.54 11538.08 66569.02 2038.75 52480.42 13822.00
29211.19
S5
7
335803.32 48226.07 6876.59 57706.79 7308.97 45886.49
30583.64
S5
8
335730.56 48152.09 6942.57 57634.28 7224.00 45815.24
30506.56
S6
7
289863.07 2610.67 52811.37 11778.10 38730.75 59.76
15519.36
S6
8
291362.45 3974.38 51311.28 13274.46 37230.76 1440.89
14027.66
S7
7
301822.19 14271.89 40858.93 23784.26 26749.88 12110.10
3442.39
S7
8
300478.63 12928.42 42200.41 22440.85 28091.65 10782.82
4780.39
The resulting distance matrix for Right hand as shown in table 8.
Table 8. Distance matrix for handgeometry images of Right
hand
Sub1 Sub2 Sub3 Sub4 Sub5 Sub6 sub7
s1
7 2827.65 289439.45 347091.15 304016.61 346337.18 305030.39 292620.94
s1
8 6749.81 293344.99 350996.60 307922.52 350243.05 308937.07 296525.46
s2
7 281275.92 5336.18 62987.84 19914.99 62234.53 20939.23 8536.34
s2
8 280999.01 5613.05 63264.75 20191.78 62511.40 21215.72 8813.11
s3
7 351933.35 65321.54 7670.70 50745.85 8434.56 49738.64 62142.12
s3
8 353353.04 66741.43 9091.93 52166.19 9857.04 51159.67 63561.14
s4
7 299450.68 12840.31 44813.59 1791.90 44061.72 2899.73 9666.54
s4
8
299450.68 12840.31 44813.59 1791.90 44061.72 2899.73 9666.54
s5
7
349192.71 62581.19 4936.45 48003.67 5683.09 46992.86 59405.81
s5
8
346047.70 59436.10 1800.04 44858.69 2538.74 43848.39 56260.68
s6
7
302660.69 16069.11 41612.90 1589.54 40853.96 472.40 12945.52
s6
8
303024.01 16432.60 41249.96 1934.73 40490.88 831.17 13308.07
s7
7
288547.06 2152.80 55725.87 12699.12 54977.75 13757.05 1295.92
s7
8
283429.18 3458.70 60852.15 17833.70 60105.59 18888.69 6411.07
In table 7 and table 8 highlighted cells indicate the
correctly classified values and highlighted values indicates the
misclassified values.
Total 32 samples are tested for handgeometry. Out of which
30 are correctly classified and 2 are misclassified. The reason for
miss-classification is the poor quality of images.

www.ijsea.com 80
Table 9. Recognition rate for Handgeometry
Test Total
no of
sample
tested
Correct
classified
Miss
classified
RR
Left
hand
16 15 1 93.75 %
Right
Hand
16 15 1 93.75 %
Total 32 30 2 93.75 %
Table 10. Overall Recognition rate for palmprint
and handgeometry
Test Total
no of
sample
tested
Correct
classified
Miss
classified
RR
Palm 56 56 0 100 %
Hand 32 30 2 93.75%
Figure 2 shows the Receiver Operating
Characteristic curve for palmprint images and Figure
3 ROC curve for handgeometry images. The ROC is
based on observed frequency and cumulative
frequency. Performance True positive rate and false
positive rates are analyzed with the ROC.
Figure 2. ROC curve for palmprint
Figure 3. ROC curve for handgeometry
5. CONCLUSION
This paper presents palm and handgeometry
recognition based on FAST features and region properties. The
results clearly indicates the significance of the method used in
this research work. For palmprint recognition the FAST
features were utilized for classification for palm into
appropriate classes. The observed classification success is
100%. The results of handgeometry recognition was improved
and achieved 93.75 % recognition. These two methods can be
collectively used for development of multimodal biometric
recognition system in very effective way.
This work is more reliable because it gives faster
results as compare to traditional biometric techniques.
For palmprint recognition the new approach FAST feature
algorithm reduces number of comparisons and provides easy
recognition rate because it gives direct discrimination between
image samples.
6. ACKNOWLEDGEMENT
We would like to acknowledge and thanks to University
Grants Commission (UGC), India for providing facility UGC
SAP (II) DRS Phase – I F. No. 3-42/2009 Biometrics:
Multimodal System Development Laboratory.
7. REFERENCES
[1] Slobodan Ribaric, Ivan Fratric, and Kristina K. 2007 A
Biometric Verification System Based on the Fusion of
Palmprint and Face Features. Department of Electrical
Engineering and Computing, University of Zagreb,
Croatia.
[2] Antonia Azzini, Ernesto Damiani, Stefania Marrara, 2007
Ensuring the identity of a user in time: a multi-modal
fuzzy approach. IEEE International Conference on

www.ijsea.com 81
Computational Intelligence for Measurement Systems and
Applications Ostuni, Italy,
[3] Teddy Ko 2005 Multimodal Biometric Identification for
Large User Population Using Fingerprint, Face and
Iris Recognition. IEEE transaction Proceedings of the
34th Applied Imagery and Pattern Recognition Workshop.
[4] V C Subbarayudu, Munaga V, N K Prasad, 2008
Multimodal biometrics system. Proceeding IEEE
transactions.
[5] Andrew Teoh, S. A. Samad and A. Hussain, 2004 Nearest
Neighbourhood Classifiers in a Bimodal Biometric
Verification System Fusion Decision Scheme. Journal of
Research and Practice in Information Technology, 36 (1).
[6] Anil K. Jain, Jianjiang Feng, Karthik Nandakumar, 2010
Biometrics : Fingerpeint Matching. Published by the
IEEE Computer Society 0018-9162 .
[7] Fan Yang, Baofeng Ma. 2007 A New Mixed-Mode
Biometrics Information Fusion Based-on Fingerprint,
Hand-geometry and Palm-print. IEEE transactions
Fourth International Conference on Image and Graphics.
[8] X. Wu, K. Wang, Zhang D. 2004 HMMs based
palmprint identification LNCS 3072, pp. 775-781.
[9] Ajay Kumar, David C. M. Wong, Helen C. Shen, Anil K.
Jain. 2004 Personal verification using Palmprint and
Hand Geometry Biometric. Department of Computer
Science, Hong Kong University of Science and
Technology, Clear Water Bay, Hong Kong.
[10] Harpreet Singh, Jyoti Raj Gulsheen Kaur , Thomas
Meitzler 2004 Image Fusion using Fuzzy Logic and
Applications. Proceeding IEEE transactions.
[11] Chun Wai Lau, Bin Ma, Helen M. Meng, Y.S. Moon and
Yeung Yam 2005 Fuzzy Logic Decision Fusion in a
Multimodal Biometric System The Chinese University of
Hong Kong, Hong Kong SAR, China.
[12] K. Ito, T. Aoki, H. Nakajima, K. Kobayashi, T. Higuchi
2006 A palmprint recognition algorithm using phase
based image matching. Proceedings of the ICIP, pp.
2669-2672.
[13] M. Wang, Q. Ruan. 2006 Palmprint recognition based
on two dimensional methods. Proceedings of ICSP.
[14] Conti, G. Milici, P. Ribino, F. Sorbello and S. Vitabile
2007 Fuzzy Fusion in Multimodal Biometric Systems.
university of 90127 Palermo, Italy.
[15] Kornelije Rabuzin, Miroslav Ba., Markus Schatten, 2008
Fuzzy Active Rules in Mutimodal Biometric Systems.
Proceedings of the ITI 2008 30th Int. Conf. on Information
Technology Interfaces.
[16] Gawande U., Sreejith R Nair, Harsha Balani, Nikhil Pawar
& Manjiri Kotpalliwar 2009 High Speed Frequency
Multimodal Biometric System Using Iris and
fingerprint. Proceeding Springer.
[17] Yong Jian Chin, Thian Song Ong, Michael K.O. Goh and
Bee Yan Hiew 2009 Integrating Palmprint and
Fingerprint for Identity Verification. IEEE Third
International Conference on Network and System Security.
[18] Cheng Lu, Di Liu*, Jisong Wang, Shuyan Wang 2009
Multimodal Biometrics Recognition by Dimensionality
Reduction Method. Second International Symposium on
Electronic Commerce and Security.
[19] Nicolas Tsapatsoulis and Constantinos Pattichis 2009
Palm geometry biometrics : A score-based fusion
Approach. Proceedings workshop AIAI.
[20] Anil K. Jain, Ajay Kumar 2010 Biometrics of Next
Generation: An Overview. proceedings SECOND
GENERATION BIOMETRICS’ SPRINGER
[21] Mohammad Imran , Ashok Rao , Hemantha Kumar G.
2011 A New Hybrid approach for Information Fusion
in Multibiometric Systems. Proceedings IEEE
transactions .
[22] Mohamed K. Shahin, Ahmed M. Badawi, and Mohamed
E. M. Rasmy 2011 Multimodal Biometric System Based
on Near-Infra-Red Dorsal Hand Geometry and
Fingerprints for Single and Whole Hands. World
Academy of Science, Engineering and Technology 56.
[23] S. Palanikumar, M. Sasikumar and J. Rajesh 2011
Palmprint Enhancement Using Discrete Curvelet
Transform. International Journal of Computer Science
Issues, 8(4), No 2.
[24] Feifei CUI, Gongping Yang 2011 Level Fusion of
Fingerprint and Finger Vein Recognition. Journal of
Computational Information Systems, 7(16) pp. 5723-5731.
[25] Romaissaa Mazouni, Abdellatif Rahmoun 2011 On
Comparing Verification Performances of Multimodal
Biometrics Fusion Techniques. International Journal of
Computer Applications (0975 – 8887), 33(7).
[26] Nishant Singh, Kamlesh Tiwari, Aditya Nigam and
Phalguni Gupta 2012 Fusion of 4-Slap Fingerprint
Images with Their Qualities for Human Recognition.
IEEE transactions IEEE .
[27] Ashutosh Kumar, Ranjan Parekh 2012 Palmprint
Recognition in Eigen-space. International Journal on
Computer Science and Engineering (IJCSE) ISSN : 0975-
3397, 4(1).
[28] P.U. Lahane, Prof. S.R. Ganorkar 2012 Efficient Iris and
Fingerprint Fusion for Person identification.
International Journal of Computer Applications (0975 –
8887), 50(17).
[29] Krishneswari, K. and S. Arumugam 2012 Multimodal
Biometrics using Feature Fusion. Journal of Computer
Science, 8 (3), 431-435.
[30] D. Y. Liliana, Eries Tri Utaminingsih 2012 The
combination of palm print and hand geometry for

www.ijsea.com 82
biometrics palm recognition. International Journal of
Video & Image Processing and Network Security
IJVIPNS-IJENS, 12(1).
[31] Gawande U. , Sreejith R Nair, Harsha Balani, Nikhil
Pawar & Manjiri Kotpalliwar 2012 A High Speed
Frequency Multimodal Biometric System Using Iris
and fingerprint. International Journal on Advanced
Computer Engineering and Communication Technology
(IJACECT) ISSN : 2278-5140, 1(2).
[32] M. P. Dale, M. A. Joshi, H. J. Galiyawala 2012 A Single
Sensor Hand Geometry and Palm Texture Fusion for
Person Identification. International Journal of Computer
Applications (0975 – 8887), 42(7).
[33] Srinivasa Rao D, Seetha M, Krishna Prasad MHM
Comparison Of Fuzzy And Neuro Fuzzy Image Fusion
Techniques And Its Applications. International Journal
of Computer Applications (0975 – 8887), 43(20).
[34] Mohamad Abdolahi, Majid Mohamadi, Mehdi Jafari, 2013
Multimodal Biometric system Fusion Using Fingerprint
and Iris with Fuzzy Logic. International Journal of Soft
Computing and Engineering (IJSCE) ISSN: 2231-2307,
2(6).
[35] Swapnali G. Garud, Apurva D. Dhawale, Mazhar Kazi,
Y.S.Rode, S.B.Dabhade, K.V.Kale, 2014 Fingerprint and
Palmprint Recognition using neighborhood operation
and FAST features. International Journal of Computer
Application (IJCA) ISBN: 973-93-80882-12-4, 95(12).
8. ANNEXURE
The KVKR Multimodal Biometrics database is designed
by research team from Biometrics Research Lab situated
in Dept. of Computer Science and IT, Dr. B.A. M.
Univeristy, Aurangabad, (M.S.) , India . under the
guidance of Prof. Dr. K.V. Kale, Head, Dept of Computer
Science and IT, Dr. B. A. M. University, Aurangabad. The
author is one member of these research team.

Palmprint and Handgeometry Recognition using FAST features and Region properties

More Related Content

What's hot (19)

Similar to Palmprint and Handgeometry Recognition using FAST features and Region properties (20)

More from Editor IJCATR (20)

Palmprint and Handgeometry Recognition using FAST features and Region properties