Biometrics: Personal Identification By Physical Attributes Of Human
Biometric is the science that deals with identification of individuals on the basis of person’s physical and behavioural attributes. It is the most advanced technology used in the modern world for identification of an individual through fingerprints, DNA or the iris. The basic perspective of biometric authentication is that every person can be accurately identified by his or her physical or behavioural traits.
Biometric system makes use of fingerprint, hand geometry, iris, retina, face, hand-vein, facial thermograms, signature or voice print to verify a person. The purpose of this paper is to convey the use of physiological attributes in terms of biometric study. The quality of a biometric system is affected by two factors; Authenticity of a sensor used, Degree of freedom offered by features extracted from sensed signals. In this paper, work on ear prints of different individuals of age 18-35 have been done and compared using various landmarks.
Review of Literature
Author Tobias Schediat et al discussed about a great perspective for fingerprint authentication using genetic algorithms. Use of fingerprint as a unique parameter was started earlier in 19th century as these were the only patterns that are constant during a human life. For experiment authors have taken fingerprints of 12 persons that is 1200 fingerprints are stored in their database and tested using different parameters for fingerprint recognition based on a genetic algorithm. They have used a new database BioGINA and primitive database FVC (Fingerprint Verification Contests) to deduce results more accurately and upgraded the systems further. New parameter tests were conducted using BioGINA and test on the databases of 2000, 2002 and 2004 were performed using FVC database.
Using BioGINA improvement reached upto 25% whereas relative improvement of equal error rate obtained was 40% of FVC 2000. Authors Arun Ross et al discussed about the problems with single biometric system and attempted to improve the performance of individual matchers by using multi biometric system. Here they have used a ‘Bimodal System’ that is produced by the fusion of two biometric systems. They have introduced three levels of fusion:
- Fusion at the feature extraction level (use to obtain only useful traits from large amount of features),
- Fusion at the matching score level (gives a matching score telling us the possibility or accuracy of the claimed identity),
- Fusion at the decision level (used to tell us about the final decision - accept or reject).
They have conducted experiments on the following physical attributes; Face Verification here they obtained a feature set of 2-D image of the prospect and matched it with template stored in database. After face detection, comparison was done by calculating the Euclidean distance between eigen face cofficients of the template and detected face. Fingerprint Verification, they used feature extraction module to calculate and obtain correct feature values. Feature values means ridge characteristics or minuatie points. Hand Geometry, they also did the experiment on hand geometry where they computed 14 features like length of fingers, width of fingers and width of palm at various locations.
The results obtained by combining the three features, but data to all the three features do not corresponds to a single set of users ; Data of face and fingerprint was obtained from user set 1 consisting of 50 users whereas the hand geometry data was collected from user set 2 comprising of 50 users. Thus, the score vector was obtained relating to the comparison and matching and represents the proximity of two feature sets as computed by a classifier. Authors have concluded the experiment by observing that, Sum rule performs better than the decision tree and linear discriminant classifiers, The benefits of multi biometrics may become even more evident in the case of larger database of users.
Authors A. Rattani et al emphasis on to study the fusion at feature extraction level for face and fingerprint biometrics. Here, they have performed some experiments to evaluate the fusion performed at feature extraction level in comparison to matching score level. The idea of putting together fusion of face and fingerprint is that face is processed as a pictorial image and fingerprint as minutiae characteristics. They have used SIFT (Scale Invariant Features Transform) to make this fusion successful.
The experiment conducted first for face recognition they employed SIFT as their basis, as this feature shows a compressed view of local grey level structure, immutable to image scaling, translation and rotation. Comparison is performed using three different techniques: Minimum pair distance, Matching eyes and mouth, Matching on a regular grid. The features are extracted using SIFT output. Fingerprint are studied using minutiae techniques where a fingerprint image is normalized, preprocessed using Gabor filters, binarized and thinned for obtaining even minute details of fingerprint. Author here emphasized on the point that with interlinking two individual features has better distinguishing capability than a individual feature. Three techniques are put together for better results. Feature set compatibility and normalization.
Now, elimination of irrelevant features was done using several feature reduction techniques. Finally, matching techniques such as point pattern matching and matching using the Delaunay triangulation technique were used and results were evaluated. The experiment result after using SIFT feature for compatibility and normalizing the minutiae points, it was really perceptible to use key descriptor as it increased the recognition for fingerprints by 1.64% and feature level of fusion by 2.64%. Lastly by using reduction technique accuracy of features increased by 0.44%. Hence, it was studied that a multimodal biometric system based fusion of two individual traits yield better results.
Authors G. Seshikal et al dispense their experiment on palm biometric with a POLY U database with discussion on other biometric parameters. Authors explained about stages of biometric systems:
- Sensor stage,
- Feature extraction stage,
- Matching stage,
- Decision stage.
Sensor stage captures the image of the feature, in next stage the feature is extracted and stored for further verification and identification whereas in further stages matching of the image is done and at last acceptance or rejection is stated according to the matching score. For their experiment authors used 600 images of 100 individuals, inner palm surface consist of different patterns that can be studied for comparison, they applied low pas filters to remove noise and then converted into binary image to remove the background. Next region of interest was specifically selected and features are extracted using multi scale edge detection and finally template matching was done and matching score was obtained.
Authors Kevin W. Bowyer et al put together their contribution to one of the four major modules of an iris biometric systems:
- Image acquisition,
- Segmentation of the iris region,
- Analysis and representation of the iris texture,
- Matching of iris representations.
First section of this paper deals with background concepts such as iris anatomy and performance of biometric system. Whereas in case of performance of biometric system they have considered four outcomes in case of verification- True accept, False accept, True reject, False reject. In case of identification- True positive, False positive, True negative, False negative. Author has summed up the contributions of different scholarly people on different subjects of iris recognition as stated above (Image acquisition, Segmentation, analysis and matching).
But, the main centre point is given to Daugman’s approach to iris recognition as. His work first gave way to iris recognition and the Daugman’s approach is the standard reference point, His substitute method of segmentation based on active contours, modifying an off angle iris image into a more frontal view, This approach is the only specific and technical one between various odds. There are various limitation in iris biometrics:
- Cataracts - natural results of aging,
- Eye injuries, certain medications and diseases,
- Glaucoma - Group of disease that reduces vision and iris biometrics remains unsuccessful due to formation of spots on person’s iris,
- Strabismus - A person a cannot align both eyes simultaneously,
- Albinism- Genetic condition results in partial or full absence of pigment.
Authors Lingyu Wang et al put forward a satisfying technique to analyze the infrared vein pattern at the back of hand for biometric purposes. Similar to fingerprints minutiae patterns (bifurcations etc.) are studied for recognition. There are average 13 minutiae points in a hand vein patterns where 7 are bifurcations and 6 are ending points. Stages of processing hand vein pattern; Image acquisition, Image Enhancement, Vein pattern segmentation, Feature extraction and matching.
The author suggests that, for evaluation of minutiae points in person’s identification he utilizes ‘Modified Hausdroff distance (MHD). They have conducted experiments have made their own FIR vein pattern image database which contains 47 district participants and comprise 141 hand vein pattern images and age group of the participants is between 18-60 years. For image acquisition they have used NEC Thermo Tracer TS7302 and the developed images are stored in 256 leveled gray-scale Bitmap (bmp) format with resolution of 320*240 pixels.
It was studied by them that the FIR images of a hand has high intensity of vein than other tissues. Next, comes the image enhancement step where the quality of the image is enhanced by further processing them by using 5*5 median filter to remove speckling noise, then they have used a 2-D wiener filter to reduce high frequency noise. After removing the speckles and reducing high frequency noise normalization methods were applied to reduce imperfections.
Lastly, image background was subtracted from veins as to get a clear image of vein patterns. Now further an skeleton of vein pattern is obtained by using skeletonization algorithm and then by applying thinning algorithm a single pixel wide skeleton of vein pattern was produced but smoothness of the image is also necessary, so to improve that a polynomial curve fitting technique was used. Now a cross number concept is carried to extract the critical points such as ridge endings and bifurcations. Result obtained was on geometrical information and it is preferred over statistical information.
Matching or comparison can be done using Hausdroff distance but at the end it was accepted that MHD measure give better results in studying minutiae of vein patterns. Experiment concluded that far infrared imaging was used to analyze the vein pattern and experiment was done by using different techniques and the experimental results show the equal error rate reaches 0% where the value of distance measured is observed to be 10. Hence, the result shows that hand vein pattern is good feature for personal identification in biometric purposes.
Author Renu Bhatia discussed about different biometric techniques such as Iris scan, retina scan and face recognition. Initially author put together some important characteristics of biometrics such as:
- Universal (characteristic lost due to accident or disease),
- Invariance of properties ( Feature should not acquire differences due to age or some sort of chronic disease),
- Measurability (Easy to capture and measure data of the physical attribute),
- Singularity (Characteristic must be unique to a person),
- Acceptance (capturing should be done on those parts which are acceptable to humans),
- Reducibillty ( Obtained data should be reduced into a file so that it could be easy to handle ),
- Reliability and tamper resistance ( Attribute used should be impossible to get modified.), Privacy (privacy of an individual should be conserved.),
- Comparable ( Trait used should be highly comparable),
- Inmitable ( No-one should reproduce the exact feature, hence less the reproducibilty more the reliability).
In section III of her paper, discussion on some types of physical attributes have been completed and most prominent topic reviewed here is face recognition. The experiment conducted is to complete facial recognition and its parameters; To procure mask of a face of individual, Use software which allows to detect the faces, Use software to compute all the size and differentiating features on the face, The face acquired is differentiated with the faces present in database, Software helps us to decide whether obtained face is a match or not.
Facial recognition contains certain parameters for comparison such as; Outline of eye sockets, Areas surrounding cheek bones, Sides of mouth and, Location of nose and eyes. In this paper the main topic studied is of face recognition as this is considered to be most fast, reliable and non- obtrusive technique. Focussing point here is that biometric is not always successful as different factors or suitable landmarks are required as to increase the efficiency and accuracy of the system.
Authors V.K Narendira Kumar et al discuss here about ear as a biometric tool and its comparison using 2D and 3D image, they have focused on multi modal biometrics using 3D ear images and compared the results with existing systems of biometric. Since, the ear growth varies with age, slight variations can be noticed during four months to eight years old and over 70 years old but ear growth is linear, after that it remains constant and due to this stability and estimate of ear growth, it is relied to be as a biometric tool. The main focus of their experiment is to compare 2D PCA- based with 3D ICP. PCA and ICP are two algorithms used. 365 people participated and out of 2342 images 823 images were selected with different conditions.
Biometric system module comprising of 4 steps: enrollment , feature extraction, matching and decision. Pre processing was done and the images acquired were normalized under 2D or 3D image. Further, landmark were selected ( Triangular Fossa and antitragus, Triangular Fossa and Incisure Intertragcia, 2 lines - one along the border between the ear and face and another from top of the ear to the bottom). Hausdroff distance is computed during the procedure and fresearcher also noticed that the 3D image data looks much cleaner than 2D image data. Various testing such as Acceptance testing, Conformity, Interoperability testing, Performance testing and Robustness testing was conducted and results were deduced as low false acceptance rate. Authors also found that most people’s left and right ear are symmetrical.
‘Pre computed Voxel closest neighbors’ strategy improved the speed of original ICP algorithm. Authors Niv Zehngut et al explore the use of nasal features as a biometric tool. They emphasize on thought that image based nose biometrics provide more recognition than full face recognition. FRGC have 3 components; a) the generic training, b) the target, c) the probe. They have conducted two experiments, in first experiment they have used FRGC (Face recognition grand challenge) matching protocols and limited probe is matched against entire unlimited target set which involves 128 million face match comparison and it was deduced that large scale nose biometrics gave better results than the size of full face.
In second experiment Nose biometrics was compared with face biometrics for occluded faces and it was concluded that nose biometrics are more significant then face biometrics. Author Himanshu Srivastava explored and compared among all kinds of biometric device available in the society and main focus is on working principal of biometric techniques and comparison on systems of biometric. Author have told about various types of biometrics under the category of Physiological biometrics and behavioral biometrics.
Author have worked and gave classification under various headings such as Biometric features used for authentication (iris , retinal, fingerprint etc), Characteristics of biometric entities ( uniqueness, permanence, universality, measureability, comparability, collect ability, invasiveness, performance, acceptability, circumvention), Social point of view ( Privacy concept, hygiene factors, safety concerns, cost factors, socially introduced, popularity, ease of use, error of incidence), technical point of view, evaluation point of view and biometric market point of view. Hence these systems are highly confidential computer based security systems and the author concluded that each by his research one can easily distinguish about the device a person should use according to the feature.
Material and Methods
20-25 Ear Samples from local population aged between 18-35 years were collected and some parameters such as length of right ear, lobule, Tragus, Helix and Anti-helix are taken into consideration. All the subjects were free from all the deformities and any sample not matching the standard was rejected. Subjects were photographed and information sheet was filled. The subjects were positioned at a distance of 1.10 m from the camera with his/her head in Frankfurt horizontal plane.
By examining these samples visually and length of the samples are taken using sliding calipers and comparison of landmarks was done. In this study distance varied from anti- helix to lobe is 5 cm to 6.1 cm, ear lobules of only 8% of samples was attached and rest of the population have hanging lobules , there is a significant difference between other parameters such as tragus have different shapes (triangular, blunt, flat, small/large), whereas helix(outer lining) and anti helix are mostly curved with different angles with bent darwin protuberance in some cases. By this experiments conclusion made that landmarks are of vital importance in biometric systems as they provide us with feature to compare between two different ear print. Usage of different landmarks have given the result that no two ears are identical.
In this work, best knowledge gained is of different biometrics systems used for specific features and different features used for biometric purposes, as author Arun Ross in his paper gave the benefits for multi biometric systems over single biometric system and gave a new characteristic by fusion of face and fingerprints. The use of SIFT feature by A. Rattani made fusion successful is really a new technique to explore. Poly U database usage by authors G. Seshikal et al for palm recognition is of great benefit for biometric purpose.
Exploring features of Daugman’s approach by author Kevin W. Bowyer et al assisted us to know about segmentation feature and use the feature specifically and technically. Far - infrared technology used by Lingyu Wang et al for studying vein patterns gave us more specific and accurate minutiae patterns and easily conclusive results. Every time biometric is not governed by techniques in some cases as discussed by authors Renu Bhatia and V.K Narendira Kumar et al that landmarks are also the points of focus for matching between two individual feature. Nose is also the best biometric tool as it comprise of various landmarks and help in biometrics as discussed by Niv Zehngut et al. The paper contains information about all the traits such as iris, nose, face , ear ,fingerprint and fusion of individual traits and studied about ear print recognition using various landmarks is of great relevance to biometric system.
- Tobias Schediat et al,”Parameter Optimization for biometric fingerprint recognition using genetic algorithms”, 2006
- Arun Ross et al , “Information Fusion In Biometrics”, 2003
- A. Rattani et al ,”Feature level fusion of face and fingerprint biometrics”
- G. Seshikal et al, “Biometric parameters & palm print recognition”, Volume 46– No. 21, May 2012
- Kevin W. Bowyer et al ,”Image understanding for iris biometrics: A survey”, Received 5 January 2007; accepted 28 August 2007 Available online 12 October 2007
- Lingyu Wang et al ,“Minutiae feature analysis for infrared hand vein patterns biometrics”, Received 12 March 2007; received in revised form 18 July 2007; accepted 19 July 2007
- Renu Bhatia,”Biometrics and Face recognition techniques”, Volume 3, Issue 5, May 2013
- Narendira Kumar et al , “Automated Human Identification Scheme using Ear Biometrics Technology” 2014
- Niv Zehngut et al, “Investigating the feasibility of image based nose biometrics”,Carnegie Mellon University, Pittsburgh, Pennsylvania 15213, USA
- Himanshu Srivastava ,”A Comparison based study on biometrics for human recognition”, 2013
- A. K. Jain, A. Ross, and S. Pankanti, “Biometric: A Tool for Information Security,” IEEE Trans. Information Forensics and Security, vol. 1, no. 2, pp. 125–144, 2006.
- E. Kukula, S. Elliott, “Implementation of Hand Geometry at Purdue University's Recreational Center”, In Proc. of 35th Annual International Carnahan Conference on Security Technology, UK, 2001, pp. 83 – 88.
- <“Biometrics and standards ITU-T Technology watch report,” 2009.
- M. A. Dabbah, W. L. Woo, and S. S. Dlay, "Secure Authentication for Face Recognition," In Proc. of IEEE Symposium on Computational Intelligence in Image and Signal Processing, USA, pp. 121 – 126, 2007.
- A. K. Jain, A. Ross, and S. Prabhakar, “An introduction to biometric recognition,” IEEE Trans. Circuits Syst.Video Technology,Special Issue Image and Video-Based Biometric, vol. 14, no. 1, pp. 4–20, 2004.