159
3C Tecnología. Glosas de innovación aplicadas a la pyme. ISSN: 2254 – 4143 Edición Especial Special Issue Marzo 2020
NEW INTUITION ON EAR AUTHENTICATION WITH
GABOR FILTER USING FUZZY VAULT
A. Kavipriya
Department of Electronics and Communication Engineering,
Kalasalingam Academy of Research and Education,
Krishnankoil, (India).
E-mail: kavidivya222@gmail.com ORCID: https://orcid.org/0000-0002-2965-2542
M. Arunachalam
Department of Electronics and Communication Engineering,
Kalasalingam Academy of Research and Education,
Krishnankoil, (India).
E-mail: muthuece.eng@gmail.com ORCID: https://orcid.org/0000-0001-8070-3475
Recepción:
05/12/2019
Aceptación:
30/12/2019
Publicación:
23/03/2020
Citación sugerida:
Kavipriya, A., y Arunachalam, M. (2020). New Intuition on Ear Authentication with Gabor Filter
Using Fuzzy Vault. 3C Tecnología. Glosas de innovación aplicadas a la pyme. Edición Especial, Marzo 2020,
159-179. http://doi.org/10.17993/3ctecno.2020.specialissue4.159-179
Suggested citation:
Kavipriya, A., & Arunachalam, M. (2020). New Intuition on Ear Authentication with Gabor Filter
Using Fuzzy Vault. 3C Tecnología. Glosas de innovación aplicadas a la pyme. Edición Especial, Marzo 2020,
159-179. http://doi.org/10.17993/3ctecno.2020.specialissue4.159-179
160 http://doi.org/10.17993/3ctecno.2020.specialissue4.159-179
3C Tecnología. Glosas de innovación aplicadas a la pyme. ISSN: 2254 – 4143 Edición Especial Special Issue Marzo 2020
ABSTRACT
At present, Frequent Biometrics Scientic Research deals with other biometric application
like Face, Iris, Voice, Hand-Based Biometrics traits for classication and spotting out the
persons. These Specic Biometric traits have their own improvement and weakness for
opting the terms like Accuracy & cost of all applications. However, in addition to other
Face-based Biometric techniques, Ear Recognition has been appealed to Boom the attention
among other Biometric researchers. This Image Template Pattern Formation of Ear cuddles
the report which is relevant for maculating the Uniqueness of their individuality. This Ear
Biometric trait observes the person’s identity based on its stable Anatomical behavior. This
biometric trait does not involve any emotional feelings with facial expressions in the same
way as a unique pair of Fingerprint. In this work, a Contemporary approach for Personal
identication is imported with Ear along with the data stores in a secured way has been
proposed. This authentication Process includes the revolution of features with Gabor Filter
and Dimension Reduction based on Multi-Manifold Discriminant Analysis (MMDA). This
work is adequately analyzed in Matlab with the Evaluation metrics such as FMR, GAR,
FNMR, by modifying the key value each time. The results of this suggested work promote
better values in recognition of individuals as for Ear modalities. Conclusively the Features
are grouped using K-Means for both identication and Verication Process. This Proposed
system is initialized with Ear Recognition Template based on Fuzzy Vault. The Key stored
in the Fuzzy Vault is utilized in safeguarding the existence of Cha Points.
KEYWORDS
2D Gabor Filter, Multi-Manifold Discriminant Analysis (MMDA), K-Means, False Matching
Rate (FMR), False Non Matching Rate (FNMR), Genuine Acceptance Rate (GAR).
161 http://doi.org/10.17993/3ctecno.2020.specialissue4.159-179
3C Tecnología. Glosas de innovación aplicadas a la pyme. ISSN: 2254 – 4143 Edición Especial Special Issue Marzo 2020
1. INTRODUCTION
In the present day scenario, the booming demand in case for both Security and automated
recognition system leads to radical research resolution in the various areas of Computer
Vision and Intelligent systems. At Present Periodic arrangements of individual identity
happens through the enactment of Password with Permissive Activities in Public security,
Access Control, Computer Vision as well as Intelligent Systems. Therefore Biometrics is
considered as a signicant application of forensics, Surveillance examination which assigns
to the technique of diagnosing the Humans by utilizing their physical or behavioral traits
along with faces, Iris, Fingerprint, Ear, Palm print, FKP, voice, and signature. These
features can be treated as Biometric diagnostic features with satisfaction of requirements:
(i)Universality, (ii)Distinctiveness, (iii)Permanence, (iv)Performance, (v)Collectability, (vi)
Acceptability. Each of these above mentioned biometric procedure has both its precedence
and nuisance using single modality which is optimal for other types of Professional systems
applications. This paper targets on Human Ear as one of the auspicious and idiosyncratic
biometric modality that involves enduring and dependent with a shape which does not
expose desperate contradiction with age.
Foseta
Helix
Antihelix
Concha
Antitragus
Lobe
Superior Crus
of Antihelix
Inferior Curs
of Antihelix
Curs of Helix
Tragus
Incisura
Figure 1. Anatomy of Human Ear.
Based on Figure 1 explains the External anatomical Structure of Human Ear with its
lingual components including Helix, Antihelix, Tagus, Concha, Lobe and further Parts.
This Innite ridges and Valleys on the external Ear’s Surface act as phonic Signals. In case
162 http://doi.org/10.17993/3ctecno.2020.specialissue4.159-179
3C Tecnología. Glosas de innovación aplicadas a la pyme. ISSN: 2254 – 4143 Edición Especial Special Issue Marzo 2020
of Low Frequencies, Pinna reproduces the acoustic signals near to the ear canal. Similarly,
High Frequencies reverberate the acoustic wave which causes the frequencies to be calm
down. This makes to note that origin of Sound perceived enabled by humans in Outer Ear.
Random Factors of Ear’s appearances can be rst examined by correlating the Left and
Right Ear of the similar Persons.
The ear Recognition can be approximately restricted based on the categories of several
methods like Hassaballah, Alshazly, and Ali (2019) explains the EAR authentication
complication utilizing Local Binary Patterns (LBP) Features. This Monotonic image of
Gray-Scale factor transforms with its computational eciency with Local Binary Pattern
features that is t for Ear Recognition problem. This tested LBP variants almost show the
accuracy rate around 99%, while the attainment faces several diculties when the level
of distortion boost. Likewise, Gandhimathi and Janarthanan (2019) describe the new
class of Biometric as Ear recognition in comparison of Fingerprint that can be smoothly
conscated from the area’s measures. Alike due to the emotion the shape of Ear does not
change even due to the emotion. It is relatively constant over a Person’s life. Robust Feature
extraction helps in determining the personality of several individuals, for instance, terrorists
at airports & stations. Similarly, Gandhimathi and Radhamani (2016), and Gandhimathi
and Janarthanan (2019), denes the eective fusion method for the combination of various
data that can be secured by the generation of cha points. These cha points help in the
formation of a secret key using the unimodal biometric data with feature vectors. The
optimal location of these feature vectors is basically created by the tness value as well
as the development of security enhancement with the help of multi-biometric systems
by means of the proposed modied template of Log Gabor Features XOR pattern.
This kind of template security is basically determined by the other way of fuzzy vault
multi- biometric cryptosystem. Vinothkanna and Amitabh (2014) explains the grouped
feature vector resemblance points that are developed by cha points and feature points.
This grouped vector points in the fuzzy vault database lead to the accurate identication
of the recognized individuals with correct match points. Several evaluation metrics like
FAR, FRR, GAR, Secret key helps in the assessment of grouped vector points. Bansal and
Hanmandlu (2017) presents the ear based identication function by the means of entropy
values with reference to change in information gain information score values. This eective
163 http://doi.org/10.17993/3ctecno.2020.specialissue4.159-179
3C Tecnología. Glosas de innovación aplicadas a la pyme. ISSN: 2254 – 4143 Edición Especial Special Issue Marzo 2020
Gaussian and Exponential function nd in generating the rened scores that are facilitated
using the Euclidean distance metrics. Anwar, Ghany, and Elmahdy (2015) developed an
advanced algorithm in case of Ear Recognition with geometric features extraction. Here
Ear recognition with geometric features extraction. The ear detection generally based on
the snake model with a median lter for removal of noises. Then canny edge and distance
metrics are created with these image features. This method is invariant to translation&
rotation with accuracy of 98%. Kacar and Kirci (2018) introduced the novel architecture
with Score Net in case of Ear recognition. This method creates with modality pool in
accordance with cascade fusion learning that is compatible with parallel processing.
Lakshman (2013) implemented the double-stage geometric approach in both scale features
& rotation invariant in case of uprooting the unique features. Hence the matching scores
are compared with the basics of threshold values with authenticating the persons. Herewith
PSO technique helps in Optimize the parameters with threshold and weights which helps in
regulating the computation time. Rathgeb, Pug, Wagner, and Busch (2016) deals with the
image compression that helps in ear recognition stages with stimulated image distortions and
partitions. Finally the detailed investigation of image compression technique. The feature
extraction was calculated with uncompressed samples of Ear databases with numerous
bitrates. Pug and Busch (2012) discovered the identication by Ear recognition in case
of 2D, 3D images in case of smart surveillance & forensic image analysis. It explains the
database collection with various features against various techniques. Nandakumar, Jain,
and Pankanti (2007) extracts the highest curvature points that are helpful in aligning the
template. Minutiae Matcher of decoding part leads to non-linear distortion which gives
a signicant improvement of GAR. Koptyra and Ogiela (2015) present a unique idea of
hiding the secrets using the fuzzy vault. It is mainly hidden the noisy data based on multi-
biometric cryptosystem. It proposes a choice of authentication accuracy relevant with a
cryptosystem on single biometric. Bae, Noh, and Kim (2003) shows the encoding of iris
code that helps in the performance of EER that gives the magnitude performance for iris
size along with processing time. Arunachalam and Kanan (2015) integrate the secret key
value using Advanced Encryption Standard to avoid several attacks like spoong, intra-
class variations, etc., for the generation of biometric key utilizing the cryptographic fusion
Uludag and Jain (2006) aims in the safeguarding and aloofness of biometric systems with the
164 http://doi.org/10.17993/3ctecno.2020.specialissue4.159-179
3C Tecnología. Glosas de innovación aplicadas a la pyme. ISSN: 2254 – 4143 Edición Especial Special Issue Marzo 2020
transformed version of the template that is stored as a cryptographic framework. So they
introduce the orientation eld of helper data for the extraction of ngerprints. Yang, Sun,
and Zhang (2011) proposed the dimensionality reduction method for pattern recognition
purposes that is based on graph embedded learning. This technique mainly based on the
construction of low dimensional data. Basically, it cannot apply for small size problem.
To overcome this, MMDA is calculated for Eigenvectors and Eigen value representations.
Yang, Sun, and Wang (2011) have attracted interest against Gabor feature with MMDA.
Certain Fuzzy vault system is generally on the support of local iris feature points from the
exact values of an unordered set with basis if shift matching technique.
Remaining paper is formulated as follows: Section 2&3 precedes the inquiry of scheduled
work and it portrayed the Fuzzy Vault which includes eradication of Ear along with Gabor
features and grouped according to k-Means clustering algorithm in a detailed manner
and Section 4 provides the particulars in relation to the basic Fuzzy Vault construction
with enrolment and verication phase. Experimental decision is essentially explained in
Section 5 and nally 6
th
section entirely organizes the basic work that provides points for
future research. The motivation of this work involves the Human Authentication that must
be considered as the most important tasks which are used in this world for the case of
identication of Persons using biometrics with its Physical and Behavioral Characteristics.
They include Fingerprints, Handprints, Palm prints, Hand veins, Eyes, Ears, Voice, and
signature. Basically, this biometric system is categorized as Unimodal; Multimodal, and
Multi-biometric system, etc. This unimodal biometric System has severe challenges
against noisy data. In this work, the Ear modalities are selected to generate the polynomial
construction to the let the secure key in a collapsed manner. The reason for selecting the
ear as main modalities is due to surprising rich features with it. Changes do not happen due
to its stable structure.
2. SYNOPSIS OF THE PROPOSED WORK
This proposed paper suggested the ow diagram that explains briey about the Ear
recognition with the generation of polynomial construction in both locking and unlocking
set for the case of Fuzzy Vault system. This work illustrates the cryptographic fuzzy vault
technique as three phases. In Recognition phase the Ear images are collected from the
165 http://doi.org/10.17993/3ctecno.2020.specialissue4.159-179
3C Tecnología. Glosas de innovación aplicadas a la pyme. ISSN: 2254 – 4143 Edición Especial Special Issue Marzo 2020
database for the further process of dimensionality reduction phase. Further the enrollment
phase the Gabor feature extraction is carried with ve scales and eight orientations. These
extracted features basically have high dimensional values. Minimization of this idea is
accessed by the MMDA Technique in the calculation of Eigen value and Eigenvector that
are explained below in this block Diagram.
RECOGNITION
IDENTIFICATION
VERIFICATION
Pre-processing
Gabor Feature
Extraction
Dimension Reduction
Techniques using MMDA
with Image shape and
Texture
Fuzzy Vault
Database
Fuzzy Vault Generation
k-Means
Clustering
Secret Key
Gabor
Feature Set
Fuzzy Vault Matcher
[Polynomial Construction
and Chaff Points
Generation]
Gabor
Feature Set
Performance MetricsSecret Key
Figure 2. Flow diagram of proposed work with Fuzzy Vault Technique.
This Figure 2 shows the process of Ear recognition with several methods that include based
on Fuzzy Vault. This vault helps in providing the security to several biometric cryptosystems.
Here the cha points are formed promptly from the biometric features which are identied
easily. The features are clustered based on manifold learning Process. The origination of
cha points or secret key by the process of the vault locking process. Locking process creates
the Polynomial generation of cha points as a key that must be entered. Similarly, the
Testing phase the same procedure is repeated in order to assess the common features and
matching is done based on the revealing of the secret key along with biometrics. There are
four signicant stages in this proposed work:
A.
Pre-processing.
B.
Gabor Feature Extraction.
166 http://doi.org/10.17993/3ctecno.2020.specialissue4.159-179
3C Tecnología. Glosas de innovación aplicadas a la pyme. ISSN: 2254 – 4143 Edición Especial Special Issue Marzo 2020
C.
Generation of Polynomial construction of grouped feature vectors.
D.
Identication and Authentication of Secret Key.
A.
Pre-Processing Phase
Since the early phase, the images are to be pre-processed along with the objective of
getting rid of the rejected part in image such as noise, Blur, reections. Originally, these Ear
images are reformed into gray scale images in the datasets are in RGB format. Thus the
training process is enforced with ear datasets. Specic basic non-linear methods recycled
are the median Filter. The main method of this lter helps in glaring of edges that helps in
reducing the noises with the point of subsiding the current pixel point with the median of
illumination in its range.
This center pixel appraisal is named as “median” and similarly the neighboring pattern as
“window”.
H(m,n)=median[x(m-k,n-l)Îw]
(1)
This equation 1 “w” imitates the window along with the pixels (m, n). Here the inured input
images of the ear are expertly pre-processed and represented as I
e
. Further this images
separable which is cropped out to obtain the ROI with the help of changing the image size
and Pixels.
B.
Gabor Feature Extraction& k-means clustering
Gabor feature Extraction is based on spatial locality and oriented selectivity with Ear
Images. Gabor wavelets formation is developed. Gabor wavelet formation is developed with
the kernels which are to identical to certain proles and exposing the desirable location and
orientation selectivity. This Gabor wavelet determination is to be entitled as:
(2)
Where u, v denotes the direction, scales of Gabor feature kernels. It is dened based on
norm operator
167 http://doi.org/10.17993/3ctecno.2020.specialissue4.159-179
3C Tecnología. Glosas de innovación aplicadas a la pyme. ISSN: 2254 – 4143 Edición Especial Special Issue Marzo 2020
Where:
(3)
This factor “K max” represents high frequency and f depicts spacing vector with ve
scales and eight orientations. Further convolution of Gabor features is based on
Z (x, y) that serves the ultimate position of the gure and *denotes convolution operator.
Multi-Manifold Discriminant Analysis (MMDA)
Collection of Sample set with various ear data label is denoted as
(4)
Likewise the linear projection of low dimensional space is dened as the
B=P
T
A
r
(5)
Considering the points with several similar class labels that Possess edge construction
between the nodes y
i
, y
j
from the corresponding class. It is also broadly promoted such as
y
i
, y
j
with its parameter
o ≤ s
ij
≤ 1 (6)
Here weight functions are taken as an important note with strict monotonically decreasing
function. Apparently, it has been noticed with negative non-symmetric that are exalted
by the matrices between Class and within class scatter in β
W
, β
B
(7)
Therefore it can be represented as:
(8)
168 http://doi.org/10.17993/3ctecno.2020.specialissue4.159-179
3C Tecnología. Glosas de innovación aplicadas a la pyme. ISSN: 2254 – 4143 Edición Especial Special Issue Marzo 2020
Hence the projection matrix is generally represented as:
β
b
p=λ β
w
p
(9)
This projection matrix is literally named after the Graph embedding algorithm which is
intended by the Eigen Value. These processes are clustered by the part of k-means clustering
by calculating the centroid points and accredited these points towards the center.
Clustering using K-means
Clustering mainly used to acclimate the feature points based on the performance of
unsupervised classication of certain patterns as groups. Considering the size of input and
classication in large groups, k-means clustering target the execution process with a basis
of Ear feature points. Further, it is continued based on centroid calculation. Basically, it is
like the expectation-maximization algorithm with mixtures of Gaussian in the process of
nding clustering with various attributes.
(10)
Where J represents the objective function that is to be dened number of cases and centroid
for cluster points that are based on Euclidean distance with distance measure dened as the
classication of objects.
Algorithm:
Input: k and other points with b
1,
b2……b; Clustering the data into several k groups.
Cluster Update: Selecting k points at random cluster Centers.
Centers Update: Assigning articles to the adjoining cluster Centre to determine according
to the Euclidean distance.
Stopping Update: Determining the centroid points or mean of severalEar featuresin ever
Cluster.
Output: Repetition of steps 2, 3 until similar points assigned to each cluster.
169 http://doi.org/10.17993/3ctecno.2020.specialissue4.159-179
3C Tecnología. Glosas de innovación aplicadas a la pyme. ISSN: 2254 – 4143 Edición Especial Special Issue Marzo 2020
c. Generation of Polynomial construction of grouped feature vectors
In order to assigning the template security, Secret key plays the main role in generating the
fuzzy vault that is united to form grouped feature vector. Originally the intake of secret
keyis concealed with the number of cha point’s generation. Considering the information
stored in the dataset is Permanent, Security is taken as an important note. Fuzzy vault is
radically a cryptographic construction recommended by Juels and Sudan (2006) securing
the critical data with the help of biometric data.
EAR DATABASE
Input
Secret Key
Fuzzy Vault
Database
Grouped Feature
Vector points
Chaff points
construction
Fuzzy Vault
Generation
Encoding
100111
0 0
100111
Figure 3. Block diagram of Fuzzy vault Construct.
Based on Figure 3 Polynomial construction with genuine points are stored as a secret key
from the Ear database. Usually, the secret key information that is distributed as unordered
sets named as Cha points. These cha points basically denote the content of secure
information to be reconstructed for revealing the secret code which is stored in the Fuzzy
vault database.
d. Identication and Authentication of Secret Key
In the recognition phase, Person’s ear images are taken as input that is pre-processed and
the features are extracted for the combination of feature vector. This input feature vector
is compared to the fuzzy vault database. Matching relates with the secret key generation
and authentication is proved. This recognition process is adorned. Let the given person’s
170 http://doi.org/10.17993/3ctecno.2020.specialissue4.159-179
3C Tecnología. Glosas de innovación aplicadas a la pyme. ISSN: 2254 – 4143 Edición Especial Special Issue Marzo 2020
Gabor feature vector that must express by c that is related to the fuzzy vault in the dataset.
In case if every feature points of the ear image matches the features in the fuzzy vault, then
the individual is admitted authentication or else the authentication is contradicted. Assured
points in the fuzzy vault will be left deserted. These points are named as “secret points”
and the x-coordinates of these features’ points provide the secret key of the authenticated
person.
Finally the procreation of the authenticated person is the second conrmation of the person
which boosts the template security.
Matching of
Feature
points
Fuzzy Vault
Database
Polynomial construction
of chaff points
Extract the features of
Ear from Gabor filter
Generation of Gabor
feature vector points
Figure 4. Recognition of a person with Fuzzy vault using Ear.
This Figure 4 represents the recognition of the person based on the person’s ear image
with Fuzzy vault. These features which are extracted from the Gabor lters are generated
to form the polynomial construction of cha points. These matched feature points are
determined from the vault database gives the authentication of the person.
3. EXPERIMENTAL RESULTS AND PERFORMANCE EVALUATION
In this category, the consequences of the designed biometric method for the recognition of
Ear modalities utilizing Fuzzy Vault are contended with detailed manner in this work. This
Intended methodology is executed in Matlab Platform of version 2017. Dataset confession
of IIT Delhi Ear images is utilized work.
171 http://doi.org/10.17993/3ctecno.2020.specialissue4.159-179
3C Tecnología. Glosas de innovación aplicadas a la pyme. ISSN: 2254 – 4143 Edición Especial Special Issue Marzo 2020
EAR-IIT Delhi
This dataset version 1.0 mainly incorporates the ear Images collected from the graduates
and Faculties at IIT Delhi, India. This entire dataset in the dataset is chiey around ages
grouped under 14-58 years. The directory of 471 images is progressively counted for every
user with unique identication number. The intention of these gures is about 272×204
pixels and reachable in jpeg format. This dataset endeavors the naturally normalized and
cropped ear images of size about 50×180 pixels beside the authentic images. A further
large adaption of ear dataset from 212 users with 754 ear images is incorporated.
Figure 5. Illustration of Ear images from IIT Delhi Ear database Version 1.0.
Experimental Results
Originally these Ear gures are in gray scale format, it is very much accessible for ltering
process. This ltering method includes Sobel lter which excludes the noise regions like
thin hair, studs etc., and the Pre-processed process these gures are shown in the Figure 6.
Figure 6. Results of Ear model (a) Input Figure (b) Preprocess Figure (c) Enhanced Image.
This Figure 6 shows the basic pre-processing and enhancement process which helps the
enhanced image after histogram equalization that further moves to feature extraction of
Gabor Filter.
172 http://doi.org/10.17993/3ctecno.2020.specialissue4.159-179
3C Tecnología. Glosas de innovación aplicadas a la pyme. ISSN: 2254 – 4143 Edición Especial Special Issue Marzo 2020
Figure 7. Magnitude and Real Parts of Gabor Filter.
The Figure 7 explains the Magnitude as well as Real Parts of Gabor feature Extraction
which is determined from the Gabor Feature Extraction.
ENTER THE KEY:
1234
SAVED
INPUT
KEY:
1234
Figure 8. Encoding Process of Fuzzy Vault.
This Figure 8 explains the input image which are grouped as feature ve0ctors has been
stored as [1234] in the Fuzzy Vault Database.
ENTER THE KEY:
1234
SAVED
INPUT
KEY: 1234
Figure 9. Decoding Process of Fuzzy Vault.
This Figure 9 shows the retrieval of Secret key from the Fuzzy Vault Database. It involves
the grouped feature points that are indulged as cha point’s generation.
Performance Evaluation Metrics
173 http://doi.org/10.17993/3ctecno.2020.specialissue4.159-179
3C Tecnología. Glosas de innovación aplicadas a la pyme. ISSN: 2254 – 4143 Edición Especial Special Issue Marzo 2020
To estimate this proposed biometric system it is established on Ear Images, several evaluation
metrics are employed. This evaluation Metrics for this work involved are False Matching
Rate (FMR), False Non-matching Rate (FNMR), Genuine Acceptance Rate (GAR) and
Accuracy.
False Matching Rate (FMR)
This Matching Rate is based on the improper recognition of un-authorized People. The
FMR is specied as resulted,
FMR = Number of un-authorized inputs with improper recognition
Total Number of Inputs
False Non Matching Rate (FNMR)
This Matching Rate is dened based on the improper recognition of authorized people. It
has been dened as designed
FNMR = Number of authorized Inputs that are falsely not recognized
Total Number of inputs
Genuine Acceptance Rate (GAR)
It is dened as the Probable of truly matching gures that are matched by the biometric
security system with the entire images in the dataset.
GAR=1-FNMR
Performance Analysis of this Proposed Work
The results of this proposed image from Ear modalities are collected from 25 samples from
various kinds of dataset. The results are taken based on the calculation of these evaluation
metrics that is explained in Table 1.
Table 1. Analysis of biometric system with Ear modality.
Sl No. FMR (%) FNMR (%) GAR (%) Accuracy (%)
1 0.62 0.40 0.60 97.0
2 0.60 0.40 0.61 98.0
3 0.63 0.40 0.62 98.8
174 http://doi.org/10.17993/3ctecno.2020.specialissue4.159-179
3C Tecnología. Glosas de innovación aplicadas a la pyme. ISSN: 2254 – 4143 Edición Especial Special Issue Marzo 2020
Sl No. FMR (%) FNMR (%) GAR (%) Accuracy (%)
4 0.64 0.40 0.60 98.8
5 0.72 0.30 0.70 96.0
6 0.71 0.30 0.70 95.0
7 0.70 0.30 0.70 98.1
8 0.70 0.30 0.71 98.2
9 0.80 0.20 0.80 96.0
10 0.90 0.10 0.90 94.1
11 0.70 0.30 0.70 98.8
12 0.90 0.10 0.70 95.5
13 0.60 0.40 0.60 94.1
14 0.50 0.50 0.50 98.2
15 0.70 0.30 0.70 98.8
16 0.80 0.20 0.80 98.3
17 0.60 0.40 0.60 94.0
18 0.60 0.40 0.60 94.2
19 0.70 0.30 0.70 95.5
20 0.50 0.50 0.50 96.1
21 0.40 0.60 0.40 97.1
22 0.70 0.30 0.70 98.1
23 0.80 0.20 0.80 98.9
24 0.90 0.10 0.90 98.3
25 0.50 0.50 0.50 98.4
Table 1 shows the performance metrics of Ear modalities biometric system with various
rates.
4. CONCLUSION
The stages in this work for this useful biometric system includes are (i) Pre-processing (ii)
Gabor Feature Extraction (iii) Polynomial Construction of grouped vector from cha
points(iv)Identication and Authentication of secret key. These Proposed work biometric
authentication systems with ear modalities are eciently implemented in Matlab. Evaluation
Metrics (FMR, FNMR, GAR) are calculated by frequently altering the key value. The
analysis of this proposed work smoothened with better accuracy value as such 98.83%.
Further this idea will involve with multimodal biometric system to check its accuracy.
175 http://doi.org/10.17993/3ctecno.2020.specialissue4.159-179
3C Tecnología. Glosas de innovación aplicadas a la pyme. ISSN: 2254 – 4143 Edición Especial Special Issue Marzo 2020
ACKNOWLEDGMENT
We would like to be grateful for the International Research Centre of Kalasalingam
Academy of Research and Education for investing nancial assistance upon the scheme
of University Research Fellowship (URF) and we also endorsed the Department of
Electronics and Communication Engineering of Kalasalingam Academy of Research and
Education, Tamil Nadu, India for providing usage of the computational facilities available
in Signal Processing and VLSI Design laboratory that were set up with the assistance of the
Department of Science and Technology (DST).
176 http://doi.org/10.17993/3ctecno.2020.specialissue4.159-179
3C Tecnología. Glosas de innovación aplicadas a la pyme. ISSN: 2254 – 4143 Edición Especial Special Issue Marzo 2020
REFERENCES
Anwar, A. S., Ghany, K. K. A., & Elmahdy, H. (2015). Human ear recognition using
geometrical features extraction. Procedia Computer Science, 65, 529-537. https://doi.
org/10.1016/j.procs.2015.09.126
Arunachalam, M., & Kanan, S. (2015). AES Based Multimodal Biometric Authentication
using Cryptographic Level Fusion with Fingerprint and Finger Knuckle Print.
The International Arab Journal of Information Technology, 12(5), 431-440. https://www.
semanticscholar.org/paper/AES-Based-Multimodal-Biometric-Authentication-
using-Arunachalam-Kannan/32a5f4d1d2f4c1d6ac521a486a7e20bc5c7b7da9
Bae, K., Noh, S., & Kim, J. (2003). Iris feature extraction using independent component
analysis. International Conference on Audio-and Video-Based Biometric Person Authentication.
Springer, Berlin, Heidelberg. https://doi.org/10.1007/3-540-44887-X_97
Bansal, M., & Hanmandlu, M. (2017). A new entropy function for feature extraction
with the rened scores as a classier for the unconstrained ear verication. Journal of
Electrical Systems and Information Technology, 4(1), 135-158. https://doi.org/10.1016/j.
jesit.2016.10.006
Gandhimathi, A., & Janarthanan, G. R. (2019). A Fuzzy vault based Multimodal
biometric cryptosystem for enhancing security. International Journal of Electrical and
Computer Engineering, 768-773.
Gandhimathi, A., & Radhamani, G. (2016). New cha point based fuzzy vault for
multimodal biometric cryptosystem using particle swarm optimization. Journal of
King Saud University-Computer and Information Sciences, 28, 381-394. https://core.ac.uk/
download/pdf/82045848.pdf
Hassaballah, M., Alshazly, H., & Ali, A. A. (2019). Ear recognition using local binary
patterns: A comparative experimental study. Expert Systems with Applications, 118, 182-
200. https://doi.org/10.1016/j.eswa.2018.10.007
177 http://doi.org/10.17993/3ctecno.2020.specialissue4.159-179
3C Tecnología. Glosas de innovación aplicadas a la pyme. ISSN: 2254 – 4143 Edición Especial Special Issue Marzo 2020
Hurley, D. J., Arbab-Zavar, B., & Nixon, M. S. (2008). The ear as a biometric. In Jain A.
K., Flynn P., Ross A. A. (eds) Handbook of Biometrics. Springer, Boston, MA, 131- 150.
https://doi.org/10.1007/978-0-387-71041-9_7
Juels, A., & Sudan, M. (2006). A fuzzy vault scheme. Designs, Codes and Cryptography, 38(2),
237-257.
Kacar, U., & Kirci, M. (2018). ScoreNet: deep cascade score level fusion for unconstrained
ear recognition. IET Biometrics, 8(2), 109-120. https://doi.org/10.1049/iet-
bmt.2018.5065
Koptyra, K., & Ogiela, M. R. (2015). Fuzzy vault schemes in multi-secret digital
steganography. In 2015 10th International Conference on Broadband and Wireless
Computing, Communication and Applications (BWCCA). IEEE. https://doi.org/10.1109/
BWCCA.2015.87
Lakshmanan, L. (2013). Ecient person authentication based on multi-level fusion of
ear scores. IET biometrics, 2(3), 97-106. https://doi.org/10.1049/iet-bmt.2012.0049
Lee, Y. J., Park, K. R., Lee, S. J., Bae, K., & Kim, J. (2008). A new method for generating
an invariant iris private key based on the fuzzy vault system. IEEE Transactions on
Systems, Man, and Cybernetics, Part B (Cybernetics), 38(5), 1302- 1313. https://doi.
org/10.1109/TSMCB.2008.927261
Nandakumar, K., Jain, A.K., & Pankanti, S. (2007). Fingerprint-based fuzzy vault:
Implementation and performance. IEEE transactions on information forensics and security,
2(4), 744-757. https://doi.org/10.1109/TIFS.2007.908165
Pug, A., & Busch, C. (2012). Ear biometrics: a survey of detection, feature extraction
and recognition methods. IET biometrics, 1(2), 114-129. https://doi.org/10.1049/iet-
bmt.2011.0003
Rathgeb, C., Pug, A., Wagner, J., & Busch, C. (2016). Eects of image compression
on ear biometrics. IET Biometrics, 5(3), 252-261. https://doi.org/10.1049/iet-
bmt.2015.0098
178 http://doi.org/10.17993/3ctecno.2020.specialissue4.159-179
3C Tecnología. Glosas de innovación aplicadas a la pyme. ISSN: 2254 – 4143 Edición Especial Special Issue Marzo 2020
Uludag, U., & Jain, A. (2006). Securing ngerprint template: Fuzzy vault with helper
data. In 2006 Conference on Computer Vision and Pattern Recognition Workshop (CVPRW’06).
IEEE. https://doi.org/10.1109/CVPRW.2006.185
Vinothkanna, R., & Amitabh, W. (2014). Fuzzy vault fusion based multimodal
biometric human recognition system with ngerprint and ear. Journal of theoretical
and applied information technology, 59(2), 304-317. http://www.jatit.org/volumes/
Vol59No2/9Vol59No2.pdf
Yang, W., Sun, C., & Wang, Z. (2011). Finger-knuckle-print recognition using Gabor
feature and MMDA. Frontiers of Electrical and Electronic Engineering in China, 6(2), 374.
https://doi.org/10.1007/s11460-011-0141-3
Yang, W., Sun, C., & Zhang, L. (2011). A multi-manifold Discriminant analysis method
for image feature extraction. Pattern Recognition, 44(8), 1649-1657. https://doi.
org/10.1016/j.patcog.2011.01.019
