Finger Knuckle-Print Authentication Using Fast Discrete Orthonormal Stockwell Transform
1st Edition
Published in January 2018
48 pages
978-3-96067-703-1 (ISBN)
Description
Biometrics refers to the authentication techniques that depend on measurable physical characteristics and behavioural characteristics to identify an individual. The biometric systems consist of different stages such as image acquisition, preprocessing, feature extraction and matching. Biometric techniques are widely used in the security world. The various types of biometric systems use different techniques for the preprocessing, feature extraction and classifiers.The dorsum of the hand is known as the finger back surface. It is highly used for personal authentication and has not yet attracted the attention of convenient researchers. It is mostly used due to contact free image acquisition. It is reported that the skin pattern on the finger-knuckle is extremely rich in texture due to skin folds and creases, and hence, can be considered as a biometric identifier. Furthermore, advantages of using Finger Knuckle Print (FKP) include rich in texture features, easily accessible, contact-less image acquisition, invariant to emotions and other behavioral aspects such as tiredness, stable features and acceptability in the society. As a result of that, there is less known use of finger knuckle pattern in commercial or civilian applications.
The local features of an enhanced palmprint image are extracted using Fast Discrete Orthonormal Stockwell Transform (FDOST). The Fourier transform of an image is obtained by increasing the scale of FDOST to infinity. The Fourier transform coefficients extracted from the palmprint image and FKP image are considered as the global information. The local and global information are physically linked by means of the framework of time frequency analysis. The global feature is exploited to refine the arrangement of FKP images in matching. The proposed schemes make use of the local and global features to verify finger knuckle-print images. The weighted average of the local and global matching distances is taken as the final matching distance of two FKP images. The investigational results indicate that the proposed works outperform the existing works.
The local features of an enhanced palmprint image are extracted using Fast Discrete Orthonormal Stockwell Transform (FDOST). The Fourier transform of an image is obtained by increasing the scale of FDOST to infinity. The Fourier transform coefficients extracted from the palmprint image and FKP image are considered as the global information. The local and global information are physically linked by means of the framework of time frequency analysis. The global feature is exploited to refine the arrangement of FKP images in matching. The proposed schemes make use of the local and global features to verify finger knuckle-print images. The weighted average of the local and global matching distances is taken as the final matching distance of two FKP images. The investigational results indicate that the proposed works outperform the existing works.
More details
Other editions
Additional editions
N. B. Mahesh Kumar | K. Premalatha
Finger Knuckle-Print Authentication Using Fast Discrete Orthonormal Stockwell Transform
Book
12/2017
1st Edition
Anchor Academic Publishing
€39.99
Shipment within 7-9 days
Persons
Dr. N.B. Mahesh Kumar is currently working as Assistant Professor (Sr.G) in Department of Computer Science and Engineering, Bannari Amman Institute of Technology, Erode, Tamil Nadu. He completed his B. Tech (IT), M.E. (CSE) and Ph.D in Anna University, Tamilnadu, India. His areas of research interest include Image Processing and Data Mining. He published 8 International Journals, 7 International Conferences, 8 National Conferences and 1 Monograph. I have 10 years teaching experience.
Content
- Finger Knuckle-Print Authentication Using Fast Discrete Orthonormal Stockwell Transform
- TABLE OF CONTENTS
- CHAPTER 1 INTRODUCTION TO BIOMETRICS
- 1.1 Introduction
- 1.1.1 Biometric Systems
- 1.2 Palmprint Biometrics
- 1.2.1 Preprocessing and ROI Extraction for Palmprint Biometrics
- 1.3 Finger knuckle-print biometrics
- 1.3.1 Finger Knuckle-print Anatomy
- 1.3.2 Preprocessing and ROI Extraction for Finger Knuckle-Print Biometrics
- 1.4 Pros of finger knuckle-print and palmprint
- 1.5 Local and Global features
- 1.6 Problem statement
- 1.7 Motivation
- 1.8 Objectives
- 1.9 Biometric Datasets
- 1.9.1 College of Engineering - Pune (COEP) Palmprint Datasets
- 1.9.2 The PolyU Palmprint Datasets
- 1.9.3 Indian Institute of Technology (IIT Delhi) Touchless Palmprint Datasets
- 1.9.4 The PolyU Finger Knuckle-Print Datasets
- 1.10 Performance Metrics
- 1.10.1 False Acceptance Rate and False Rejection Rate
- 1.10.2 Speed
- 1.10.3 Equal Error Rate (EER)
- 1.10.4 Correct Classification Rate (CCR)
- 1.10.5 Data Presentation Curves
- 1.10.5.1 Receiver Operating Characteristic (ROC) Curve
- CHAPTER 2 FINGER KNUCKLE-PRINT IDENTIFICATION BASED ON LOCAL AND GLOBAL FEATURE EXTRACTION USING FAST DISCRETE ORTHONORMAL STOCKWELL TRANSFORM
- 2.1 Overview of Fast Discrete orthonormal Stockwell transform
- 2.2 Local - Global Feature Extraction and Matching
- 2.2.1 Local Feature
- 2.2.2 Global Feature
- 2.3 Local global information fusion for knuckle-print recognition
- 2.4 Experimental results and discussion
- 2.5 Summary
- CHAPTER 3 CONCLUSIONS AND FUTURE WORK
- 3.1 SUMMARY AND CONCLUSIONS
- 3.2 FUTURE WORKS
- REFERENCES
Text Sample:
Chapter 1.9 Biometric Datasets
1.9.1 College of Engineering - Pune (COEP) Palmprint Datasets
The COEP palmprint database (COEP Palm Print Database (College of Engineering Pune) 2010) consists of 8 different images of single person's palm. The database consists of total 1344 images pertaining to 168 persons. The dataset is collected over a period of one year. The images were captured using digital camera. The resolution of images is 1600×1200 pixels.
1.9.2 The PolyU Palmprint Datasets
The PolyU Palmprint Database (Zhang 2010) contains 7752 gray scale images corresponding to 386 different palms in BMP (Bitmap) image format. Twenty samples are collected from each of these palms in two sessions. Each 10 samples were captured in the first session and the session, correspondingly. The average intermission among the first and the second collection is two months period of time.
1.9.3 Indian Institute of Technology (IIT Delhi) Touchless Palmprint Datasets
The IIT-Delhi palmprint image database (Kumar 2007) consists of the hand images collected from the students and staff at IIT-Delhi, India. This dataset is acquired in the IIT Delhi campus during July 2006 - Jun 2007 using a simple and touchless imaging setup. The images are collected in the indoor atmosphere and employ circular fluorescent illumination around the camera lens. The presently accessible dataset is from 235 users. All the subjects in the database are in the age group 12-57 years. In each subject, seven images are collected from each of the left and right hand. All the images were collected in fluctuating hand posture differences. All the subjects are offered with live feedback to present the person's hand in the imaging region. The touchless imaging consequences in higher image scale variations. The resolution of these images is 800 × 600 pixels and all these images are available in bitmap format. Finally 150 × 150 pixels are automatically cropped and normalized palmprint images are also available.
1.9.4 The PolyU Finger Knuckle-Print Datasets
PolyU FKP database (Zhang 2009) consists of 7920 images collected from 660 different fingers. The samples are collected in two separate sessions. In each session; six images are collected for the left index and left middle finger, the right index and right middle finger. From each person, 48 images are collected from 4 fingers. The size of the acquired FKP images is 768×576 under resolution above 400 dpi. Based on the experiments, high resolution images are not necessary for feature extraction and pattern matching. Therefore, Gaussian smoothing operation is applied to the original image. The smoothen image is down sampled to about 150 dpi. Hence the size of ROI images is 110×220 pixels.
1.10 Performance Metrics
Performance testing comprises a critical aspect of biometric modality assessments. Investigators are able to draw from a wide range of performance evaluation metrics that assess functional system accuracy and usability. The choice of metrics employed in performance testing is considered by the type of biometric modality or system undergoing evaluation precisely, whether the scheme is traditional in nature (i.e. a well-established, single transaction identification modality such as Fingerprint, Face, or Iris recognition) or novel in nature (e.g. an emerging modality such as Pulse, or an innovative application such as cognitive biometrics).
Traditional performance metrics describe system accuracy, precision and usability. The ability of an authentication system to measure a biometric with a high degree of closeness to the biometrics' true value is known as accuracy. The repeatability of accurate system measurements over time is known as precision. The ease with which a system used is termed as usability. The majority of traditional biometric performance metrics derives from signal detection theory. It seeks to quantify the ability to discern between information-bearing energy patterns (signals) and the random energy patterns (noise) that obstruct the informative pattern detection and acquisition. Traditional biometric performance metrics are referred and applied in a variety of ways, taking into consideration: performance evaluation type (technical, scenario, or operational testing), performance component assessment (detection, acquisition, enrollment, matching, and authentication), human factors (usability), and others.
The performance of a biometric feature, result or application is distinguished by dissimilar metrics. The user needs to enroll his biometric traits when the biometric system is used for the first time. The biometric scheme needs palmprints, finger knuckle-print from the operator. This input is stored in the database as a template. It is internally linked to a User ID (Identification). The biometric input is matched with the templates in the database by a pattern matching algorithm when the user wants to authenticate or identify the person for the first time.
Performance metrics generally take the system of rates for each metric. It is important to note that the measured/observed rate in any evaluation is distinct from the predicted/expected rate that occurs in deployed, fully operational biometric systems (predicted/expected performance rates may be gauged using measured/observed rates).
1.10.1 False Acceptance Rate and False Rejection Rate
The probability that the system incorrectly authorizes an unauthorized individual due to wrongly matching the biometric input with a template is known as False Acceptance Rate. The FAR is normally expressed as a percentage of invalid inputs which are incorrectly accepted. False Accept Rate is also called as False Match Rate.
The false acceptance rate describes the proportion of identification or verification transactions in which an impostor subject is incorrectly matched to a genuine user template stored within a biometric system. FAR reflects the ability of a non-authorized user to access a system, whether via zero-effort access attempts or deliberate spoofing or the other methods of circumvention.
The probability that the system incorrectly rejects the access to an authorized individual due to deteriorating the wrong match is known as the false rejection rate. The FRR is normally expressed as a percentage of valid inputs which are incorrectly rejected. FAR and FRR are much dependent in the biometric factor and the technical implementation of the biometric solution. A personal FRR is determined for each individual because FRR is purely person dependent. Take this into account while determining the FRR of a biometric solution; one person is insufficient to establish an overall FRR for a solution. Due to environmental conditions or incorrect use, for example when using dirty palmprints on a palmprint reader, the FRR is increased. Mostly the FRR lowers when a user gains more experience in using the biometric device or software. False Reject Rate is sometimes referred as False Non-Match Rate.
The FRR describes the proportion of identification or verification transactions in which a genuine subject is incorrectly rejected from a biometric system. FRR may occur as a result of user presentation error or the exploitation of formerly enrolled authentication templates.
FAR and FRR are the key metrics for biometric solutions, certain biometric devices or software even allow to tune them so that the system more quickly matches or rejects. Both FRR and FAR are significant, but for greatest applications one of them is considered as most important. Two examples to exemplify this:
1. When biometrics is used for logical or physical admittance control, the objective of the application is to prohibit access to unauthorized individuals in all situations. It is strong that a very low FAR is desirable for such an application, even if it comes at the cost of a higher FRR.
2. When surveillance cameras are used to screen a crowd of people for missing children, the objective of the application is to identify any missing offspring that come up on the screen. When the identification of those offspring is computerised using a face recognition software, this software has to be set up with a low FRR. As such a greater quantity of matches will be false positives, but these are studied rapidly by surveillance personnel.
1.10.2 Speed
The time taken to enroll the people in the template and the time taken by an individual to be authenticated is given by the manufacturers of biometric devices and software.
1.10.3 Equal Error Rate (EER)
The equal error rate describes the point at which genuine and imposter error rates are closest to zero. EER can be represented as a percentage with time/unit factors (e.g. results of "8.3% EER for 1sec/1heartbeat" in a Pulse modality study). EER is not beneficial in considering definite scheme performance, but can be helpful as a first-order performance indicator for 1:1 verification systems.
Chapter 1.9 Biometric Datasets
1.9.1 College of Engineering - Pune (COEP) Palmprint Datasets
The COEP palmprint database (COEP Palm Print Database (College of Engineering Pune) 2010) consists of 8 different images of single person's palm. The database consists of total 1344 images pertaining to 168 persons. The dataset is collected over a period of one year. The images were captured using digital camera. The resolution of images is 1600×1200 pixels.
1.9.2 The PolyU Palmprint Datasets
The PolyU Palmprint Database (Zhang 2010) contains 7752 gray scale images corresponding to 386 different palms in BMP (Bitmap) image format. Twenty samples are collected from each of these palms in two sessions. Each 10 samples were captured in the first session and the session, correspondingly. The average intermission among the first and the second collection is two months period of time.
1.9.3 Indian Institute of Technology (IIT Delhi) Touchless Palmprint Datasets
The IIT-Delhi palmprint image database (Kumar 2007) consists of the hand images collected from the students and staff at IIT-Delhi, India. This dataset is acquired in the IIT Delhi campus during July 2006 - Jun 2007 using a simple and touchless imaging setup. The images are collected in the indoor atmosphere and employ circular fluorescent illumination around the camera lens. The presently accessible dataset is from 235 users. All the subjects in the database are in the age group 12-57 years. In each subject, seven images are collected from each of the left and right hand. All the images were collected in fluctuating hand posture differences. All the subjects are offered with live feedback to present the person's hand in the imaging region. The touchless imaging consequences in higher image scale variations. The resolution of these images is 800 × 600 pixels and all these images are available in bitmap format. Finally 150 × 150 pixels are automatically cropped and normalized palmprint images are also available.
1.9.4 The PolyU Finger Knuckle-Print Datasets
PolyU FKP database (Zhang 2009) consists of 7920 images collected from 660 different fingers. The samples are collected in two separate sessions. In each session; six images are collected for the left index and left middle finger, the right index and right middle finger. From each person, 48 images are collected from 4 fingers. The size of the acquired FKP images is 768×576 under resolution above 400 dpi. Based on the experiments, high resolution images are not necessary for feature extraction and pattern matching. Therefore, Gaussian smoothing operation is applied to the original image. The smoothen image is down sampled to about 150 dpi. Hence the size of ROI images is 110×220 pixels.
1.10 Performance Metrics
Performance testing comprises a critical aspect of biometric modality assessments. Investigators are able to draw from a wide range of performance evaluation metrics that assess functional system accuracy and usability. The choice of metrics employed in performance testing is considered by the type of biometric modality or system undergoing evaluation precisely, whether the scheme is traditional in nature (i.e. a well-established, single transaction identification modality such as Fingerprint, Face, or Iris recognition) or novel in nature (e.g. an emerging modality such as Pulse, or an innovative application such as cognitive biometrics).
Traditional performance metrics describe system accuracy, precision and usability. The ability of an authentication system to measure a biometric with a high degree of closeness to the biometrics' true value is known as accuracy. The repeatability of accurate system measurements over time is known as precision. The ease with which a system used is termed as usability. The majority of traditional biometric performance metrics derives from signal detection theory. It seeks to quantify the ability to discern between information-bearing energy patterns (signals) and the random energy patterns (noise) that obstruct the informative pattern detection and acquisition. Traditional biometric performance metrics are referred and applied in a variety of ways, taking into consideration: performance evaluation type (technical, scenario, or operational testing), performance component assessment (detection, acquisition, enrollment, matching, and authentication), human factors (usability), and others.
The performance of a biometric feature, result or application is distinguished by dissimilar metrics. The user needs to enroll his biometric traits when the biometric system is used for the first time. The biometric scheme needs palmprints, finger knuckle-print from the operator. This input is stored in the database as a template. It is internally linked to a User ID (Identification). The biometric input is matched with the templates in the database by a pattern matching algorithm when the user wants to authenticate or identify the person for the first time.
Performance metrics generally take the system of rates for each metric. It is important to note that the measured/observed rate in any evaluation is distinct from the predicted/expected rate that occurs in deployed, fully operational biometric systems (predicted/expected performance rates may be gauged using measured/observed rates).
1.10.1 False Acceptance Rate and False Rejection Rate
The probability that the system incorrectly authorizes an unauthorized individual due to wrongly matching the biometric input with a template is known as False Acceptance Rate. The FAR is normally expressed as a percentage of invalid inputs which are incorrectly accepted. False Accept Rate is also called as False Match Rate.
The false acceptance rate describes the proportion of identification or verification transactions in which an impostor subject is incorrectly matched to a genuine user template stored within a biometric system. FAR reflects the ability of a non-authorized user to access a system, whether via zero-effort access attempts or deliberate spoofing or the other methods of circumvention.
The probability that the system incorrectly rejects the access to an authorized individual due to deteriorating the wrong match is known as the false rejection rate. The FRR is normally expressed as a percentage of valid inputs which are incorrectly rejected. FAR and FRR are much dependent in the biometric factor and the technical implementation of the biometric solution. A personal FRR is determined for each individual because FRR is purely person dependent. Take this into account while determining the FRR of a biometric solution; one person is insufficient to establish an overall FRR for a solution. Due to environmental conditions or incorrect use, for example when using dirty palmprints on a palmprint reader, the FRR is increased. Mostly the FRR lowers when a user gains more experience in using the biometric device or software. False Reject Rate is sometimes referred as False Non-Match Rate.
The FRR describes the proportion of identification or verification transactions in which a genuine subject is incorrectly rejected from a biometric system. FRR may occur as a result of user presentation error or the exploitation of formerly enrolled authentication templates.
FAR and FRR are the key metrics for biometric solutions, certain biometric devices or software even allow to tune them so that the system more quickly matches or rejects. Both FRR and FAR are significant, but for greatest applications one of them is considered as most important. Two examples to exemplify this:
1. When biometrics is used for logical or physical admittance control, the objective of the application is to prohibit access to unauthorized individuals in all situations. It is strong that a very low FAR is desirable for such an application, even if it comes at the cost of a higher FRR.
2. When surveillance cameras are used to screen a crowd of people for missing children, the objective of the application is to identify any missing offspring that come up on the screen. When the identification of those offspring is computerised using a face recognition software, this software has to be set up with a low FRR. As such a greater quantity of matches will be false positives, but these are studied rapidly by surveillance personnel.
1.10.2 Speed
The time taken to enroll the people in the template and the time taken by an individual to be authenticated is given by the manufacturers of biometric devices and software.
1.10.3 Equal Error Rate (EER)
The equal error rate describes the point at which genuine and imposter error rates are closest to zero. EER can be represented as a percentage with time/unit factors (e.g. results of "8.3% EER for 1sec/1heartbeat" in a Pulse modality study). EER is not beneficial in considering definite scheme performance, but can be helpful as a first-order performance indicator for 1:1 verification systems.
