IOSR Journal of Engineering (IOSRJEN) www.iosrjen.

org
ISSN (e): 2250-3021, ISSN (p): 2278-8719
Vol. 08, Issue 4 (April. 2018), ||VI|| PP 54-62

Multimodal Biometric Authentication System Using Face and

Fingerprint Biometric
Jasmeen Sharma1, Dharam Veer Sharma2
1
(DepartmentOf Computer Science, Punjabi University, India)
2
(Department Of Computer Science, Punjabi University, India)
Corresponding auther:Jasmeen Sharma

ABSTRACT:Biometric Authentication Systems Are Used Widely As A Part Of Many Softwares Or

Applications. These Are Of Generally Two Types: Unimodal Biometric System And Multimodal Biometric
System. Unimodal Uses One Biometrics And Multimodal Uses More Than One Biometrics. Unimodal Systems
Are Less Preferred Because Of Physiological Defects. Here, Multimodal Biometric System Is Proposed Using
Two Biometric: Face And Fingerprint. In The Proposed System, Principal Component Analysis (PCA),
Bacterial Foraging Optimization Algorithm (BFOA), Minutiae Extraction And Multi-Layer Neural Network
(MLNN) Are Used. Minutiae Points Are Extracted From Fingerprint Images. Facial Features Are Extracted
Using PCA. These Features Are Optimized Using BFOA Separately. The Selected Features Are Then Fused
Using Weighted Sum Rule. Then The System Is Trained Using MLNN And It Works As Classifier. Then
Testing Is Performed By Following The Same Procedure. System Generates Message As Genuine/Match Or
Imposter/Not Match. This Proposed System Is Tested Using 200 Sample Images. This Proposed System Is
Evaluated Using Three Parameters: False Acceptance Rate (FAR), False Rejection Rate (FRR) And Genuine
Acceptance Rate (GAR). It Shows Accuracy As 99%.

KEYWORD:Bacterial Foraging Optimization Algorithm, Minutiae Extraction, Multilayer Neural Network,

Principal Component Analysis.
----------------------------------------------------------------------------------------------------------------------------- ----------
Date of Submission: 23-03-2018 Date of acceptance: 07-04-2018
--------------------------------------------------------------------------------------------------------------------------------------

I. INTRODUCTION
The Area That Deals With The Kinds Of Signals For Which The Input Is A Picture And The Output Is
Also Defined As Image Is Completed In Image Processing. It Can Be Further Separated Into Two Forms I.E.
Analog And Digital Image Processing. Digital Image Processing Has Number Of Merits As Compared To
Analog Image Processing. There Are Numerous Algorithms Which Can Be Used Over Digital Images Rather
Than Analog Images. Digital Image Processing Makes The Understanding Easy While Dealing With Image.
A Picture Might Be Considered As A Two Dimensional (X,Y) Area, Where X And Y Are Spatial
Directions, And The Adequacy Or Value At Any Combination Of Directions (X, Y) Is Known As The Fixation
Or Dark Level Or Pixel Value Of The Picture. Pixel Is The Smallest Element In A Picture. Pixel Is Itself
Signified As “Picture Element”. Whenever X, Y, And The Focus Estimations Of An Image Are For The
Limited Part Or Discrete Amount, We Call The Picture A Computerized Picture. AComputerized Picture Is
Made Up Of A Limited Number Of Components, Each Of Which Has A Specific Place And Esteem. These
Components Are Called Picture Components, Buddies, And Pixels. Pixel Is The Term Utilized Most Broadly To
Mean The Components Of A Computerized Picture. It Is Not Wrong To Say That Pictures Play The Most
Essential Part In Human Mindfulness. Even For People Who Are Lacking To The Visual Band Of The
Electromagnetic (EM) Range, Imaging Contraption Cover Nearly The Whole EM Range, Going From Gamma
To Radio Waves. They Can Work On Pictures Produced By Source That People Are Not Comfortable To
Connect With. These Include Ultrasound, Electron Microscopy, And PC Produced Pictures. Now And Then A
Refinement Is Made By Characterizing Picture Handling As A Control In Which Both The Information And
Yield Of A Procedure Are Pictures [1].

Multimodal Biometric System:Biometric System Is A System Which Identifies Or Authenticates The Person
Based On Biometric Traits. BiometricAre The Unique Features Or Characteristics Of The Individual. Biometric
System Can Be Used For Variety Of Purposes Like Access Control, Person Identification, Attendance In
Various Departments Etc. It Is A Pattern Recognition System That Gives Results By Acquiring Biometric
Information From An Individual, Extracts The Feature-Set From The Acquired Data And Calculates The
Characteristic Set Against The Pattern Set In The Database. Biometric Systems Can Use One Or More
International organization of Scientific Research54 | P a g e
 Multimodal Biometric Authentication System using Face and Fingerprint Biometric

Biometrics Characteristics For The Purpose. For Instance, Uni-Modal Biometric System Works On One
Biometric Like On Fingerprints Or On Facial Expressions Etc. On The Other Hand Multi-Modal Biometric
System Works On More Than One Unique Feature Of A Human Being Like Both On Fingerprint And Facial
Expressions. Basically Biometric Are Of Two Types:
1. Physiological
2. Behavioural
Physiological Biometrics Are Those Which Are Related To Physical Parts Of Human Body Like Iris,
Fingerprints, Face, Ear, Veins, Hand Palm Etc. On The Other Hand, Behavioural Biometrics Are Those That
Deal With Day To Day Activities Of Human. They Deal With Behaviour Of A Human Being. This Is Emerging
Technology. As Many Researches Show That Every Person Is Unique In His Behaviour. Behavioural
Biometrics Includes Way Of Talking, Way Of Walking, Way Of Typing, Our Gestures, Keystrokes, Different
Expressions Etc.Multi Modal Biometrics Systems Make Use Of More Than One Biometrics Identifiers.
Biometric Identifier Can Be Both Physiological, Both Behavioral Or The Combination Of Them. . They Have
So Many Advantages Over Uni-Modal Systems. . It Is Often Used Where High Security Is Needed.One Trait
Can Be Fooled But Fooling All The Traits Is Next To Impossible.

II. LITERATURE SURVEY

Reddy Et Al. [2] Proposed A Multimodal Biometric Framework Based On PCA, LBP And PNN
(Probabilistic Neural Network). The Unimodal Biometric Framework Is Least Used Because Of Its
Physiological Defects. In Proposed Strategy LBP Extracted The Face Highlights From Confront Pictures And
Those Highlights Are Given As Contribution To PCA That Produces Face Feature Vector With Decreased
Dimensions. Finger Highlights Are Extracted From Fingerprint Pictures Utilizing LBP And Those Highlights
Are Given As Contribution To PCA That Creates Finger Feature Vector With Lessened Dimensions. Utilizing
LBP, The Unmistakable Literary Highlights Of Face And Unique Mark Are Extricated. Weighted Summation
Fusion Technique Is Utilized To Consolidate These Features/ Highlights. A Probabilistic Neural System Is
Utilized As Classifier. A Normal Acknowledgment Rate Of 97.5% Accomplished With Proposed Strategy.
Kanade Et Al. [3] Proposed An Idea Of Feature Level Fusion With Weighted Error Correction To
Obtain A Multi-Biometric Feature Vector Which Is Used To Get A Secure Template. In This Proposed Work, A
Cryptographic Key Is Generated. Biometrics Needs Revocability And Protection While Cryptography Can't
Identify The Client's Character. By Getting Cryptographic Keys Utilizing Biometrics, One Can Accomplish The
Properties, For Example, Revocability, Confirmation About Client's Character, And Protection. As The Left
And Right Irises Are Not Correlated, One Can Take Them As Two Autonomous Biometric. The Watchword
Enhances Revocability, Protection, And Security Of The Framework.
Kumar And Shekhar [4] Proposed An Approach For Personal Recognition Using Rank-Level
Combination Of Multiple Biometrics Representations. Non-Linear Rank Level Fusion System Is Designed And
Compared With Rank Level Fusion System. Not Many Efforts Are Required To Study Rank Level Fusion For
Multi Biometric Combination. The Comparative Experimental Results From The Publicly Available Multi
Biometrics Scores And Real Hand Biometrics Data To Evaluate/Ascertain The Rank-Level Combination Using
Borda Count, Logistic Regression/Weighted Borda Count, Highest Rank Method, And Bucklin Method Are
Presented. An Experimented Result Suggests That Essential Performance Enhancement In Recognition And
Accuracy Achieved As Compared To Individual Palm Print Representations. They Also Suggested That
Proposed Nonlinear Rank-Level Approach Supersedes The Rank-Level Combination Approaches.
Rathgeb Et Al. [5] Proposed A System That Generates An Irreversible Representation Of Multiple
Biometric Templates. These Are Developed Using Adaptive Bloom Filters. The Proposed Work Uses Two
Identifiers (Face And Iris) For Fusion And Generates Single Protected Template. It Helps In Improving The
Privacy As Compared To Single Biometrics System.
Sanjekar Et Al. [6] Discussed Merits Of Multimodal Biometrics. Uni Modal Biometric Systems Have
Many Demerits. Noisy Data, Inter Class Similarities, Intra Class Variation And Spoofing Affects The
Performance Of The System. Moreover, Unimodal System Is Less Accurate And Secure Than Multi-Modal
System. That’s Why Multimodal Systems Are Preferred. Multiple Biometric Identifiers Are Used In Multi-
Modal Biometric System For Person Authentication. This Paper Explains The Basic Multi-Modal System, Its
Working And Phases Of Fusion Through The Help Of Block Diagram.
Jagadiswary Et Al. [7] Proposed A Fused Multimodal System. It Has Various Benefits As Compared
To Uni-Biometric Framework Like Improved Verification Accuracy, Bigger Space To Coordinate With More
Examples. Biometric Authentication Systems Are Just An Add-On Of Pattern Recognition Framework. Optical
Sensors Such As Scanning Devices And Cameras Are Used To Record Images And Unique Features. The Aim
Is To Minimise The FAR (False Acceptance Rate) And Maximise The GAR (Genuine Acceptance Rate). The
Proposed And Improved Multimodal Confirmation Framework Depends Upon Feature Extraction By Using

International organization of Scientific Research55 | P a g e

 Multimodal Biometric Authentication System using Face and Fingerprint Biometric

Retina, Fingerprint, Etc. And Key Generation. MATLAB Was Used As A Framework. The Performance Of
Framework Improved As Genuine Acceptance Rate Of 95.3% And False Acceptance Rate Of 0.01%.
Singh Et Al. [8] Developeda New Biometric Framework Using Ear And Iris, Which Was Initially
Created By Edged Detection Proceeded By Withdrawing Features By Using PCA Technique. It Was
Continuously Developing And Used In Various Applications. Multiple Biometric Systems Are A Blend Of Two
Or More Uni-Modal Biometric Systems. Multimodal Biometric Systems Are Furnished For Using 2 Or More
Physiological Or Behavioural Identifiers.
Dinca Et Al. [9] Presented A Review On Multi-Biometric Systems, Including Fusion Methods And
Security. Combination Is A Main Prerequisite In Multi-Biometric Frameworks, Being The Technique Used To
Consolidate Numerous Biometric Strategies Into A Framework. The Combination Segment Overviews The
Strategy Of Combining The Biometric. The Security Area Deals With The Current Issues, For Example: Format
Security, Sensor Satirizing, And Biometric Encryption. Contextual-Based Biometric Are Also Discussed.
Kim Et Al. [10] Reviewed Several Multi-Biometric Techniques Along With Fusion Of Biometrics.
Biometrics Become Most Encouraging Technologies In Previous Few Years That Utilized Physiological
Features Like Face, Voice, Fingerprint, Iris, Etc. For Individual Identification. Combination Of Two Or More
Techniques Provides Better Performance As Compared To Unimodal Systems. Lastly, Discussed Few
Applications Of Smart TV Domain Depend Upon Multimodal Biometric.
Omran Et Al. [11] Proposed A Multimodal System Using Iris And Finger-Print Recognition
Framework. Fingerprint Identification Formula Is Revised To Create Delaney Triangulation Framework In
Which Neighboring Triangles Were Compared Among Stored Templates And Input. Whereas Iris Recognition
System Was Revised Segmentation Method Based On Correlation Filter. Such Method Was Applied To Lower
Part Of Iris Region Which Is Supposed To Be Least Affected By Noise. The Suggested Multimodal System
Provides High Accuracy And Less Error Rate Close To (0.9%).
Kumar Et Al. [12] Discussed Currently Used Multimodal Biometric Confirmation. Authentication Is
A Process To Validate Person Identity Via Specific Inputs. Authentication Becomes Hot Topic Of Research
Because Of Attacks On Computer Networks. This Paper Aims At To Describe The Best Combinations Of
Biometric Characteristics For Authentication. It Also Mentioned The Merits And Demerits Of Biometric
Mechanisms To Improve The Performance.
Awalkar Et Al. [13] Devised An Algorithm Which Uses The Combination Of Iris And Face. Iris Is
Chosen For Its Good Recognition Power. Face Is Easily Available And Can Be Captured Easily. Score Level
Fusion Is Used For Fusing The Match Scores. LBP (Local Binary Patterns) And Gabor Filters Are Used For
Extracting Features From Face Images And Daugman’s Algorithm Is Used For Iris Feature Extraction. The
Performance Was Evaluated And EER (Expected Error Rate) Was Turned Out To Be 1.48%.
Dhriti Gupta [14] Reviewed Different Levels Of Fusion In Biometric Systems. It Is The Need Of The
Hour To Provide Security To A Person And Property. So, The Technology Is Developing In That Way.
Biometric Systems Do The Needful. They Provide The Required Security. These Can Be Uni-Modal Or Multi-
Modal Depending On The Number Of Identifiers They Use. The Comparison Is Made On The Basis Of Min-
Max, Tan H Normalization And Z-Score Techniques. The Results Showed That Multi-Modal Biometric
Systems Are Far Better Than Uni-Modal Systems In Dealing With Noisy Images, Spoofing, Error-Rates And
Non-Universality.
Shende Et Al. [15]Discussed The Benefits Of Fingerprint And Iris Recognition Systems. These Can
Be Used As Detection Techniques For Classifying The Client As Genuine Or Fake. Biometric Detection System
Is More Reliable And Accurate Than Other System Such As Card, Password, Etc. Capable To Give Good
Performance And Better Protection Against Spoofing Attacks.

III. MINUTIAE EXTRACTION

Minutiae Based Extraction Is More Fruitful Than Based On Fingerprints Directly. There Are Local
Ridge Features Existing As A Part Of Fingerprints Which Are Called Minutiae. The Results Are More Accurate
Using Minutiae Features. In Case Of Fingerprints Matching, Every Time You Need To Match The Fingerprints
Against The Recorded Fingerprints Templates Of Every User In The Database. It Results In Lot Of
Computations And Search. But This Is Not The Case With Minutiae. Though Templates Are Also Used In
Minutiae Based System But The Size Is Relatively Small. Moreover Minutiae Points Ensure Uniqueness. But It
Too Depends On The Quality Of Image. A Good Quality Image May Have 30-90 Minutiae Points. There Can
Be Many Types Of Minutiae But Two Are Main: Ridge Endings And Bifurcation. It Is Not Like That The
Fingerprints Only Have These Minutiae Features. There Are Other Features Too But Minutiae Features Have
Orientation Maps Which Supersede All Other Features. They Are Robust Too. Using Minutiae Is Like Reducing
Complexity Of The Problem From Complex Fingerprint Recognition To Pattern Matching. Finger Prints Are
Permanent And Unique Feature Of Human Body. Fingerprints Though Are Prone To External Environment
Like With Age They Become Less Recognizable. But Minutiae Features Remains There Forever. We Can Say

International organization of Scientific Research56 | P a g e

 Multimodal Biometric Authentication System using Face and Fingerprint Biometric

Minutiae Are More Stable. Moreover There Is No Scope Of Making Copy Of Minutiae Points As Is The Case
With Fingerprints. So, Minutiae Extraction Handles The Privacy Issue In Better Way. Yeah, Poor Quality Image
Is A Challenge For Minutiae Extraction. For That One Can Use Image Enhancement Techniques [16].
Minutiae Points Can Be Achieved With High Accuracy Using Image Segmentation Techniques. Image
Segmentation Is Required To Separate The Foreground From Noisy Image. Noisy Mage Leads To Inaccurate
Minutiae Points. For Keeping The Ridge Flow Genuine, Image Enhancement Techniques Can Be Used.

 Ridge Ending:Ending Of The Ridge Flow.

Minutiae Points Can Be Of Various Types As Listed Below.

 Bridges:It Denotes That Ridge Which Joins Two Ridges.

 Ridge Islands:It Denotes The Ridge Which Is Surrounded By Other Ridges.
 Ridge Bifurcation:It Denotes The Point Where One Ridge Divides Into Two.
 Ridge Dots:These Are Small Ridges Of Similar Size To A Dot.
 Ponds / Lakes:These Are Blank Spaces Among Diverging Ridges.
 Crossovers:It Denotes The Point Where Two Ridges Cross Each Other. [17]

Figure 1: Minutiae Points

IV. BACTERIAL FORAGING OPTIMIZATION ALGORITHM

BFOA Is An Algorithm Used Globally For Optimization Of Problems And Inspired From Behaviour
Of Enzymes Namely Escherichia Coli. It Is Invented By Passino. It Is Encouraged From Chemo Taxis
Behaviour Of Bacteria. E. Coli Bacteria Search For Nutrients In Such A Way To Get Maximum Energy In A
Unit Time. The Movement Of Bacteria In Search Of Nutrients Is Termed As Chemo Taxis. When They Get
Sufficient Nutrients, They Become Lengthier In Size And Break Down Into Two Bacteria. This Activity Of
Bacteria Is Termed As Reproduction. Due To Some Reasons Like Environment Changes, Some Bacteria Dies
And Some Shifted To Another Part Or Group, Which Results In The Constancy Of The Population. This
Behaviour Of Bacteria Can Be Termed As Elimination-Dispersal Activity. All These Activities Are Also
Depicted In The Algorithm. So, Here, We Can Say The Said Algorithm Consists Of Four Main Processes As

 Chemo Taxis:In His Process, Algorithm Depicts The Movement Of Cells. E. Coli. Cells Can Move Using
Depicted Below:

Two Modes: Swimming And Tumbling. Swimming Is Moving In Same Direction And Tumbling Denotes

 Swarming:Group Behaviour Has Been Observed In E. Coli Bacteria To Form Patterns Or Swarms. They
The Movement In Random Direction. Cells Can Alternate Between The Two In Their Lifetime.

 Reproduction:The Least Healthy (Or Highest Cost In Terms Of Its Application) Dies And Some Healthier
Form Groups Or Rings While Travelling Through The Medium.

 Elimination And Dispersal:As Described Above As Well, Changes In The Living Environment Of The
Bacteria Breaks Down Into Two. The Population Size Hence Remains Intact.

Bacteria Results In The Killing Of A Group Of Bacteria Or Shifting Of A Group Of Bacteria To Other
Place. To Implement This In Algorithm, A Probability Is Used To Eliminate Some Features And Some Are
Replaced In The Search Space.

Steps Of BFOA
Parameters: The Parameters Used In The Algorithm Are As Follows:

International organization of Scientific Research57 | P a g e

 Multimodal Biometric Authentication System using Face and Fingerprint Biometric



D: Dimension Of The Search Space


T: Total Number Of Bacteria


Zc: Number Of Chemotactic Steps


Zs: Swimming Length


Zr: Number Of Reproduction Steps


Zed: Number Of Elimination- Dispersal Steps


Ped: Elimination- Dispersal Probability
S(I): Size Of Step Taken Randomly (Tumble)
Let P(A,B,C)= { Θi (A,B,C) , Where I= 1,2,…,T} Represent The Position Of Each Member In The
Population At Ath Chemotactic Step, Bth Reproduction Step And Cth Elimination-Dispersal Step. Let J(I,A,B,C)
Denotes The Cost At The Location Of The Ith Bacterium Θi (A,B,C) ∈Rd (Rational Numbers With D
Dimensions)
Step 1: Initialise Parameters D, T, Zc, Zr, Zs, Zed, Ped, S(I) (I=1,2,…,T), Θi
Step 2: Elimination-Dispersal Loop: C=C+1
Step 3: Reproduction Loop: B=B+1
Step4: Chemo Taxis Loop: A=A+1
1. For I=1,2,…,T Repeat The Below
2. Compute Fitness Function J(I,A,B,C)
Let J(I,A,B,C)= J(I,A,B,C) +Jcc((Θi (A,B,C), P(A,B,C))
3. Let Jlast(I,A,B,C)=J(I,A,B,C) Until We Find The Best One
4. Tumble: A Random Number Vector Is Generated ∆ (I) ∈Rd With Each Element Lies In[-1,1]
5. Take A Chemotactic Step As Follows:
Θi (A+1,B,C)= Θi (A,B,C)+S(I) ∆ (I)/ (Sqrt(∆T(I) ∆ (I)))
6. Calculate J(I,A+1,B,C) Using
J(I,A+1,B,C)= J(I,A,B,C) +Jcc((Θi (A+1,B,C), P(A+1,B,C))
7. Swim:
A. Let G=0 (Swim Length Counter)
B. While G<Zs Repeat
C. Let G=G+1
D. If J(I,A+1,B,C)<Jlast, Let Jlast= J(I,A+1,B,C) And
Θi (A+1,B,C)= Θi (A,B,C)+S(I) ∆ (I)/ (Sqrt(∆T(I) ∆ (I)))
And Use This To Find J(I,A+1,B,C)
E. Else G=Zs. End Of While Loop
8.Go To Next Bacterium I=I+1, I!=T
Step 5: If A<Zc, Repeat Step 4 As The Bacteria Life Is Not Over
Step 6: Reproduction:
1. For The Given B And C, And For Each I=1,2,…,T Let
Ji Health= ∑ J=1,2,…,Zc+1 J(I,A,B,C)
Be The Health Of Bacteria. More Heath Signifies Less Cost. Arrange Them In Ascending Order Of
Jhealth.
2. The Tr Bacteria With The Highest Jhealth Values Die And Tr Best Bacteria Split Into Two.
Step 7: If B<Zr Go To Step 3 Till The Reproduction Steps Are Over.
Step8: Elimination-Dispersal: For I=1,2,…,T With Probability Ped, Eliminate And Disperse Each Bacterium I
Till C<Zed.

V. PRINCIPAL COMPONENT ANALYSIS

It Is A Method Of Analysing Arrangement In Data, And Expressing The Information In Such A Way
As To Focus Their Resemblance And Difference [18]. PCA Converts The Set Of Observations Of Correlated
Variables Into A Set Of Values Of Linearly Uncorrelated Variables By Using Orthogonal Transformation.
These Set Of Values Are Called Principal Components. It Is One Of The Simplest Techniques Of Analysis That
Uses Eigen Values And Eigen Vectors. It Can Be Thought Of As A Technique To Explain The Internal
Structure Of The Data Set Or Features In A Way That Makes It Easier To Solve The Issue Or The Problem. It
Describes How Much Data Is Varying. It Is Mostly Used For Reducing Dimension Of The Image So That The
Required Information Is Obtained And Extra Information Can Be Discarded. The Aim Is To Select A Good
Quantity Of Images In Order To Have The Best Understanding Of Problem With The Minimum Database.
Below Are The Steps Which Are Followed In PCA.
Step 1: First Create The Data Set Or Feature Set Or Feature Vector.
Step 2: Mean Is Calculated For The Vectors.
Step 3: Mean Is Subtracted From Each Value.
International organization of Scientific Research58 | P a g e
 Multimodal Biometric Authentication System using Face and Fingerprint Biometric

Step 4: Co-Variance Is Calculated Of The Results Of Step 3. A Co-Variance Matrix Is Obtained.

Step 5: Form The Above Covariance Matrix, Eigenvalues And Eigen Vectors Are Calculated.
Step 6: Eigen Values And The Eigen Vectors Are The Principal Components.

VI. IMPLEMENTATION
Step 1: Data Collection: In This Multimodal Biometric Fusion Data Is Collected From The UCI Machine
Learning Repository Site. First Dataset Contains Face Images And Second One Contains Finger Print Images.
Step 2: Conversion:The Multi Model Biometric System Needs To Convert The Original To Gray Scale Image.
This System Works On Gray Scale Images.

Figure 2: Proposed Flow Chart

Step 3: Feature Extraction In Fingerprint: Mostly, Local Ridge Features Are Used For Identification
Systems Based On Fingerprints. These Local Ridge Features Are Called Minutiae. Here, In This Proposed
System Minutiae Points Are Extracted From Fingerprints. Minutiae Points Are Extracted Using Binary And
Thinned Image. First, Binary Image Is Extracted From Gray Scale Image And Then Resultant Binary Image Is
Processed To Have A Thinned Image. From That Thinned Image, Minutiae Points Are Extracted. These Will Be
Used For Matching.
Step 4: Feature Extraction In PCA (Principle Component Analysis):Principal Component Analysis
Algorithm Is Used To Extract Features From Face Images. This Algorithm Can Be Used For Dimension
Reduction, Data Analysis And Feature Extraction. It Defines The Eigen Values And Eigen Vectors.
Step 5: Feature Selection Using BFOA Algorithm:Large Number Of Features Are Extracted In The Previous
Steps (Step 3 And Step 4). Such A Large Feature Set Adds To Cost So It Is Beneficial To Optimize The Feature
Set. By Optimising It, We Mean To Select Some Of The Features Which Are More Prominent Among Them
And Use Them In Next Steps For Identification. For That Purpose We Use BFOA I.E. Bacterial Foraging
Optimization Algorithm. The Whole Step By Step Information Of The Working Of It Is Explained In Previous
Chapters.

International organization of Scientific Research59 | P a g e

 Multimodal Biometric Authentication System using Face and Fingerprint Biometric

Step 6: Fusion:Fusion Of Multiple Biometric Traits Is Required For Multi-Biometric Authentication System.
This Gives More Accurate Results In Recognition Of A Person. Here, Weighted Sum Rule Based Technique Is
Used For Fusion. Fusion Is Basically Done To Get A Single Vector Or A Scalar Number From Different
Feature Sets Of The Different Modalities. It Is Performed While Training As Well As Testing Phases.
Step 7: Multi-Layer Neural Network:In This Classification Method, MLNN Is Used To Train The Features
And Test The Features. There Are Three Layers Apparently Exist In MLNN Namely Input Layer, Hidden Layer
And Output Layer. Except For The Input Features Each Feature Is A Neuron That Uses A Non-Linear
Activation Function Using Sigmoid Method. MLNN Uses A Supervised Method For Training And For Testing
Method, Simulation Model Is Used To Analyse The Similar Features. The Result Comes Out To Be True
Identity /Match Exists Or An Imposter Or Not A Match. According To Which Access Can Be Granted Or
Denied.

 False Acceptance Rate (FAR): It Defines How Many Users Are Falsely Accepted As True Identities. It Is
Step 8:Compute Performance Parameters:

 False Rejection Rate (FRR): It Is The Measurement Of How Many Genuine Users Rejected Falsely. It
The Measurement For Selection Of Imposters As Genuine Persons.

 Genuine Acceptance Rate (GAR): It Is The Opposite Of FRR. It Defines The Accuracy Of The System Or
Gives The Probability Of Rejecting Genuine Persons Considered As Imposters.

It Gives The Probability Of Selecting A Genuine User As A Genuine One.

GAR= 1- FRR.

VII. DATABASE DESCRIPTION

Face Dataset:Face Images Files Are In PGM Format. The Size Of Individual Face Image Is 92*112 Pixels With
256 Gray Level Per Pixel.The Images Are Defined In 40 Directories, Which Have Names Of The Form XV,
Where X Indicates The Subject Number. Each Subject Has 10 Images Having Different Expressions, Poses Etc.
Each Of These 10 Dissimilar Images Are Defined By Y [19].

Figure 3: Face Images

Fingerprint Dataset:Fingerprints Are Taken From CASIA Unique Finger Impression Picture Database. It Has
20000 Images For 500 Subjects Or Individuals. Every Individual Contributes For 40 Images Of His Fingers.
Each Image Is Taken By Putting Different Levels Of Pressure On The Sensor. Each Image Is Of Size 328*356
[20].

Figure 4: Fingerprint Images

VIII. CONCLUSION AND FUTURE SCOPE

Conclusion: In The Conclusion, Multi-Modal Biometric System Is Compared To Uni-Modal Biometrics. It Is
Now In-Use For Few Applications. It Has Implemented A Multi-Modal Biometric System Based On PCA
(Principle Component Analysis) And Minutiae Algorithm With MLNN Classification And BFOA Methods That
International organization of Scientific Research60 | P a g e
 Multimodal Biometric Authentication System using Face and Fingerprint Biometric

Will Use The Facial And Fingerprint For Verification Of The Person. In This Work, Bacteria Foraging
Optimization Method Is Used To Select The Feature Based On The Extracted Features. It Is The Intension Of
This Research Technique To Calculate The Fusion. In This Research Work, PCA (Principle Component
Analysis) And Minutiae Algorithm Are Used For Feature Extraction Using Facial Images And Fingerprint
Images Respectively. Principle Component Analysis Can Be Used For Reducing Dimensions, Data Analysis
And Feature Extraction. Fingerprint Images Features Are Extracted From Fingerprint Images Using Minutiae
Algorithm. Optimization Is Done Through BFOA (Bacteria Foraging Optimization) That Evaluates Feature
Vector Of The Features With Reduced Dimensions. The Different Texture Features Of Facial And Fingerprint
Are Extracted.Weighted sumFusion Method Is Used To Connect These Uni-Modal Features. It Could Conclude
That The Fusion Produces Better Identification Consequences When Compared With Single Modalities. The
Performance Study Using Database Might Be Performed And Calculate The Performance Parameters Like
GAR, FAR And FRR. In This Proposed Work, GAR Or Accuracy Is 99%. False Rejection Rate Is 1%. False
Acceptance Rate Is 1.3%.

Future Scope:ThisWork Can Be Extended With Other Biometric Identifiers. Other Biometrics Like Iris Can
Also Be Added To It With Already Existed Biometrics. This Can Be Used As A Part Of Other Systems Too
Like For Security Purpose. Other Algorithms Can Be Used To Improve Its Performance.

REFERENCES
[1]. Gonzalez, R.C. “Digital Image Processing”, Pearson Education India, 2nd Edition, Pp 1-28, 2009.
[2]. Reddy Ch Venkata Rami, KV Krishna Kishore, U. Srinivasulu Reddy, And M. Suneetha. "Person
Identification System Using Feature Level Fusion Of Multi-Biometrics." International Conference
On Computational Intelligence And Computing Research (ICCIC), IEEE, Pp. 1-6, 2016.
[3]. Kanade Sanjay, Dijana Petrovska-Delacrétaz And Bernadette Dorizzi, "Multi-Biometrics Based
Cryptographic Key Regeneration Scheme.", In Biometrics: Theory, Applications, And Systems,
BTAS'09, IEEE 3rd International Conference, Pp. 1-7, 2009.
[4]. Kumar Ajay, And Sumit Shekhar. "Personal Identification Using Multibiometrics Rank-Level Fusion.",
IEEE Transactions On Systems, Man, And Cybernetics, Part C (Applications And Reviews), Vol 41, No.
5, Pp 743-752, 2011.
[5]. Rathgeb Christian, And Christoph Busch, "Cancelable Multi-Biometrics: Mixing Iris-Codes Based On
Adaptive Bloom Filters.", Computers And Security, Vol. 42, Pp 1-12, 2014.
[6]. Sanjekar P. S. And J. B. Patil, "An Overview Of Multimodal Biometrics." Signal And Image Processing,
Vol 4, No. 1, Pp 57, 2013.
[7]. Jagadiswary D., And D. Saraswady, "Biometric Authentication Using Fused Multimodal
Biometric.", Procedia Computer Science, Vol. 85, Pp 109-116, 2016.
[8]. Singh Paramjeet And Er Rohini Sharma, "To Improve The Security In Biometric Authentication Based
On Multi-Model Fusion System.", International Journal Of Applied Sciences And Tehnology, Vol. 2,
Issue No. 4, Pp 198-202, 2017..
[9]. Dinca Lavinia Mihaela And Gerhard Petrus Hancke. "The Fall Of One, The Rise Of Many: A Survey On
Multi-Biometric Fusion Methods.”, IEEE Access 5, Pp 6247-6289, 2017.
[10]. Kim Yeong Gon, Kwang Yong Shin, Won Oh Lee, Kang Ryoung Park, Eui Chul Lee, Cheonin Oh, And
Hankyu Lee, "Multimodal Biometric Systems And Its Application In Smart TV.", Computer Applications
For Database, Education, And Ubiquitous Computing, . Springer, Berlin, Heidelberg, Pp. 219-226, 2012.
[11]. Omran Safaa S., And Maryam A. Salih., "Design And Implementation Of Multi-Model Biometric
Identification System." International Journal Of Computer Applications (0975–8887) Vol. 99, No. 15, Pp
14-21, 2014.
[12]. Kumar K. And Mohammed F., "A Review Of Multimodal Biometric Authentication Systems",
International Journal Of Scientific And Technology Research, Vol. 5, No. 12, Pp. 5-9, 2016.
[13]. Awalkar Kirti V., Sanjay G. Kanade, Dattatray V. Jadhav, And Pawan K. Ajmera. "A Multi-Modal And
Multi-Algorithmic Biometric System Combining Iris And Face." International Conference
On Information Processing (ICIP), IEEE, Pp. 496-501, 2015.
[14]. Gupta Dhriti, "Multimodel Biometric System: Fusion Techniques And Their Comparison." ,
2ndInternational Conference On Recent Advances In Engineering &Computational Sciences (RAECS),
IEEE, Pp. 1-4, 2015.
[15]. Shende Pradnya M. And Dr MV Sarode, "Multiple Biometric System Application: Iris And Fingerprint
Recognition System." International Journal Of Application Or Innovation In Engineering &Management
(IJAIEM), Vol. 5, Issue 3, 2016.
[16]. Bansal Roli, Priti Sehgal, And Punam Bedi. "Minutiae Extraction From Fingerprint Images-A Review.",
International Journal Of Computer Science Issues, Vol. 8, Issue 5, No. 3, Pp 74-85, 2011.

International organization of Scientific Research61 | P a g e

 Multimodal Biometric Authentication System using Face and Fingerprint Biometric

[17]. Barbu Tudor, Adrian Ciobanu And Mihaela Luca. "Multimodal Biometric Authentication Based On
Voice, Face And Iris.", E-Health And Bioengineering Conference (EHB), IEEE, Pp. 1-4, 2015.
[18]. Https://Www.Bayometric.Com/Minutiae-Based-Extraction-Fingerprint-Recognition/
[19]. Http://Www.Cl.Cam.Ac.Uk/Research/DTG/Attarchive:Pub/Data/Att_Faces.Zip
[20]. Http://Biometrics.Idealtest.Org/Detailsdatabase.Do?Id=3.

Jasmeen Sharma"Multimodal Biometric Authentication System Using Face And Fingerprint

Biometric" IOSR Journal of Engineering (IOSRJEN), vol. 8, no. 4, 2018, pp. 54-62.

International organization of Scientific Research62 | P a g e