Gender Classification based on Fingerprints using SVM

Romany F. Mansour

1,2

, Abdulsamad Al-Marghilnai

, Meshrif Alruily

Department of Computer Science, Faculty of Science, Northern Border University, Arar, Saudi Arabia

Department of Science and Mathematics, Faculty of Education, N.V., Assiut University, Assiut, Egypt

College of Computer Science & Information, Northen Border University, Arar, Saudi Arabia,

College of Computer Science & Information, Aljuof University, Skaka, Saudi Arabia

Key

words: Fingerprint, Gender Classification, SVM, Biometrics.

Abstract: The fingerprint is commonly used biometric method for person identification. It is the most conventional

and widely used technique in forensics and criminalities. Identification of the person's age and gender based

on his/her fingerprint is an important step in overall person's identification. The aim of this research paper is

to propose a gender classification technique based on fingerprint characteristics of individuals using discrete

cosine transform (DCT). Gender classification evaluated using dimensionality reduction techniques such as

Principal Component Analysis (PCA), along with Support Vector Machine (SVM). A dataset of 2600

persons of different ages and sex was collected as internal database. Of the samples tested, 1250 samples of

1375 exactly identified male samples and 1085 samples of 1225 exactly identified female samples.

1 INTRODUCTION

The fingerprint is commonly used biometric method

for person identification. It is the most conventional

and widely used technique in forensics and

criminalities. Identification of the person's age and

gender based on his/her fingerprint is an important

step in overall person's identification; this area still

needs more work (IBG 2007; Kralik M. & Novotny

V.2003; Hall J. & Kimura D.1994; Karine C.et al

2000 and Acree & Mark A.,1999). (Badawi et al

2006) indicated that gender classification is the most

important step in forensic anthropology that can be

used to reduce the list of suspects and reduce the

search domain.

There is evidence that male and females

fingerprint characteristics - such as ridge count,

thickness and density - are different. Accra showed

that females have a higher ridge density (Acree,

Mark A.,1999 and

Gungadin S 2007 ) while Kralik

showed that the males have higher ridge breadth

(Hall J. & Kimura D.1994). Also, researchers

showed that both males and females have higher

rightward directional asymmetry in the ridge count

(Karine C.et al 2000 and Acree & Mark A.,1999;

Badawi et al 2006 and Sanders G.& Kadam A.2001)

with the asymmetry being higher in males than

females (Austin R.et al 2001) . Figure 1 shows an

example of two different fingerprints for a male and

female.

(Gnanaswami et al., 2012 ) proposed a method

based on discrete wavelet transform (DWT) and

singular value decomposition (SVD). K nearest

neighbor (KNN) is used as a classifier and overall

correct classification rate of 88% is achieved. (Ritu

et.al 2012) used frequency domain analysis for

fingerprint based gender identification. The

classification is obtained by analyzing fingerprints

using Fast Fourier transform (FFT) and other

techniques. Tom, et al, (Rijo et al 2013)] has

proposed a technique based on frequency domain

analysis to estimate gender. They achieved an

overall accuracy rate of 70%.

The aim of this research paper is to propose a

gender classification technique based on fingerprint

characteristics of the individual. In section 2

research approach is discussed, results are presented

in section 3 and we conclude in section 4.

2 RESEARCH APPROACH

2.1 The Major Feature

The gender of each person can be learned from the

features of their fingerprints. General features

241

F. Mansour R., Al-Marghilnai A. and Alruily M..

Gender Classiﬁcation based on Fingerprints using SVM.

DOI: 10.5220/0004721602410244

In Proceedings of the 6th International Conference on Agents and Artiﬁcial Intelligence (ICAART-2014), pages 241-244

ISBN: 978-989-758-015-4

 2014 SCITEPRESS (Science and Technology Publications, Lda.)

Figure 1: Example of Two different fingerprints for a male

and female: (a) Two different fingerprints for a male

showing no (or few) white lines and small RTVTR and (b)

Two different fingerprints for a female showing large

count of white lines and large RTVTR (Badawi et al

2006).

include RTVTR, fingerprint pattern type, line count,

and the type of pattern matching between

corresponding left and right tracks. The average

ratio between the ridge thickness and the valley

thickness for each of the fingerprints is computed

and an average ratio is computed for every person.

For each fingerprint the white lines count and

ridge count are extracted manually; then, the average

white lines count as well as the ridge count was

calculated for each subject. Pattern type is extracted

manually for each fingerprint. The pattern type

concordance was calculated for the fingerprints of

each right-left corresponding fingerprint pair for the

subject (1 if the corresponding fingerprints have the

The ridge count asymmetry between the right-

left corresponding fingerprints of a person is

calculated. The asymmetry is 1 for a left-right

corresponding fingerprint pair if the ridge count of

the left fingerprint is greater than the right one, is -1

if it is smaller, and is 0 if both ridge counts are

equal.The pre-classification steps include image

enhancement, noise reduction, binarization and

extraction of features; same approach used by

(Girgis et al 2009)followed.

2.2 Classification Methods

The frequency domain approach is adopted where

the DCT is used to analyze figure print properties

and obtain its vectors of coefficient. Once a database

of fingerprint is obtained, PCA is used to reduce

data dimensionality and identify sets of unique

characteristics are called the principal components.

The most significant m vectors are then chosen on

the basis of the eigenvalues. The value of m is

chosen by considering the cumulative sum of the

eigenvalues. The features of an image x are then

computed by projecting it into the space spanned by

the eigenvectors. These feature vectors are used

during training and classification. Support vector

machine (SVM) is used to classify the subjects

according to their gender. SVM results are compared

with Fisher linear discriminant and quadratic

discriminant function results (Kirby M. & Sirovich

L.,1990; Moghaddam, B.& Yang M.2000 and

Belhumeur V.et al 1997).

3 RESULTS

Two databases were used for testing the

performance of the gender classification system. The

databases were named as ‘synthetic’ and ‘internal

dataset’. The first was synthetically generated

database (DB4) from the FVC2006 competition

(Jayadevan R.et al 2006). The second was the

database in which the Fingerprint samples were

scanned from 2600 persons of different ages and

gender (1375 males, and 1225 females) were

obtained from different places that used biometric

fingerprint sensor for marking the attendance and

were analyzed using frequency domain analysis. The

images in the entire two databases had a size of

240x320 pixels and have a resolution of 500dpi. The

developed algorithm has been tested using the MAT

LAB 7.1.

3.2 Compute RTVTR

Measuring the Ridge thickness to valley thickness

ratio (RTVTR), the following results were getting

for 20 randomly selected samples. The result shows

that the females have a higher RTVTR compared to

the males as shown in figures 2 and 3.

3.3 Compute Ridge Count

Ridge count is the number of ridges occurred in a

particular region of a particular section of the

fingerprint. The result of the Ridge count is shown

in the table 1 below, and it shows that the males

have a slightly higher ridge count than the females.

From the results above, three observations can

rightly be made firstly the females have a higher

ICAART2014-InternationalConferenceonAgentsandArtificialIntelligence

242

Figure 2: A histogram of the RTVTR obtained for

females.

Figure 3: A histogram of the RTVTR obtained for males.

ridge thickness to valley thickness ratio than the

males, secondly the Males has a slightly higher ridge

count than the females. And there is no particular

relationship between the age of subjects and their

fingerprint pattern, as it does not change (only as a

result of accident or mutation). Table 2 shows the

comparison between SVM and other classifiers

techniques.

Table 2: Comparison between SVM and other classifiers

techniques.

Classifier

Error Rate

Overall Male Female

SVM 10.2 % 9.1 % 11.4 %

FLD 50.7 % 49.4 % 52.4 %

Quadratic classifier 40. 9 % 36.7 % 44.8 %

Table 1: The result of the ridge count.

# image Males Females

1 14.642 13.661

2 14.352 13.781

3 14.253 12.978

4 13.948 13.465

5 14.645 13.875

6 16.473 13.667

7 14.731 13.657

8 14.532 13.898

9 14.572 13.675

10 14.493 13.643

11 14.343 13.794

12 14.637 13.103

13 15.362 13.133

14 14.546 12.981

15 14.691 13.408

16 15.356 13.675

17 14.572 13.223

18 14.478 13.454

19 14.398 13.107

20 14.604 14.134

4 CONCLUSIONS

The aim of this research paper was to propose a

gender classification technique based on fingerprint

characteristics of individuals using discrete cosine

transform (DCT). Gender classification evaluated

using dimensionality reduction techniques such as

Principal Component Analysis (PCA), along with

Support Vector Machine (SVM). To classify, we

extracted the most significant features based DCT

existing database. These features were used to train

the SVM classifier. The experimental results showed

that the proposed system can be used as a primary

candidate in forensic anthropology with an accuracy

of 96.39%. For DB4, and from the internal database,

1,375 samples were tested 1,225 men and women

samples. The optimal threshold for each

transformation is chosen for best results. It is found

that SVM produces an accurate decision about 92 %

for women and 76 % for men. SVM provides greater

accuracy compared to other existing techniques.

REFERENCES

Acree, Mark A.,1999, "Is there a gender difference in

fingerprint ridge density?." Forensic science

international 102.1, 35-44.

GenderClassificationbasedonFingerprintsusingSVM

243

Austin R., Christopher L., 2001, “Implications of the

IDENT/IAFIS Image Quality Study for Visa

Fingerprint Processing”, Mitertek Systems (MTS).

Badawi A., Mahfouz M., Tadross R., and Jantz R., 2006,

"Fingerprint-Based Gender Classification," In Proc. of

the International Conference on Image Processing,

Computer Vision, Pattern Recognition, Las Vegas,

Nevada, USA, Vol. 1.

Belhumeur V., Hespanha J., and Kriegman D., 1997,”

Eigenfaces vs.fisherfaces: Recognition using class

specific linear projection”, Transactions on Pattern

Analysis and Machine Intelligence, PAMI19 (7): 711–

720.

Girgis M., Sewisy A., Mansour R., 2009, "A robust

method for partial deformed fingerprints verification

using genetic algorithm", Expert Systems with

Applications 36, PP. 2008–2016.

Gnanasivam .P, and Dr. Muttan S,2012, “Gender

Identification Using Fingerprint through Frequency

Domain analysis” IJCSI International Journal of

Computer Science Issues, Vol. 9, Issue 2, No 3.

Gungadin S. 2007, “Sex Determination from Fingerprint

Ridge Density”. Internet Journal of Medical;2(2):4-7.

Hall J. and Kimura D.1994, Dermatoglyphic asymmetry

and Sexual Orientation in Men. Behavioral

Neuroscience, 108, 1203-1206.

IBG, 2007. Biometrics Market and Industry Report. IBG:

NY.

Jayadevan R., Jayant Kulkarni V., Suresh N. Mali,Hemant

K.,2006,“A New Ridge Orientation based method for

feature extraction from fingerprint images, ”

Proceedings of World Academy of Science,

Engineering and Technology, Volume 13.

Karine C., Christopher M. and Martin L.2000, “Birth

Order, Birth Interval, and Deviant Sexual Preferences

among Sex Offenders.” Sexual Abuse: A Journal of

Research and Treatment, Vol. 4, No. 1.

Kirby M., Sirovich L., 1990,"Application of the

Karhunen-Loeve procedure for the characterization of

human faces” IEEE Trans. Pattern. Anal Machine

Intelligence, vol. 12, no. 1, pp. 103-108.

Kralik, M., Novotny V. 2003,” Epidermal ridge breadth:

an indicator of age and sex in paleodermatoglyphics”,

Variability and Evolution, Vol. 11: 5–30.

Moghaddam, B.; Yang M., 2000, ”Gender classification

with support vector machines. In: Automatic Face and

Gesture Recognition”, 2000. Proceedings. Fourth

IEEE International Conference on. IEEE, p. 306-311.

Rijo J., Arulkumaran T., 2013 “Fingerprint Based Gender

Classification Using 2D Discrete Wavelet Transforms

and Principal Component Analysis”. International

Journal of Engineering Trends and Technology,

Volume 4 Issue 2.

Ritu K. and Susmita G., 2012, “Fingerprint Based Gender

Identification using Frequency Domain Analysis”.

International Journal of Advances in Engineering &

Technology.

Sanders G., Kadam A. 2001, “Prepubescent children

show the adult relationship between dermatoglyphic

asymmetry and performance on sexually dimorphic

tasks.”, Cortex., 37(1):91-100.

ICAART2014-InternationalConferenceonAgentsandArtificialIntelligence

244