Search

For face recognition task the PCA plus LDA technique is a famous two-phrase framework to deal with high dimensional space and singular cases. In this paper, we examine the theory of this framework: (1) LDA can still fail even after PCA procedure. (2) Some small principal components that might be essential for classification are thrown away after PCA step. (3) The null space of the within-class scatter matrix Sw contains discriminative information for classification. To eliminate these deficiencies of the PCA plus LDA method we thus develop a new framework by introducing an inverse Fisher criterion and adding a constrain in PCA procedure so that the singularity phenomenon will not occur. Experiment results suggest that this new approach works well. [ABSTRACT FROM AUTHOR]

In this paper, a new iris segmentation method for Hand-held capture device is proposed. First, the pupil is binarized using the intensity threshold, then use morphologic method to denoise the eyelashes and eyelids noise. The geometrical method is used to calculate the coordinates of the pupil. Second, the outer (or limbus) boundary is localized using the shrunk image with the Hough transform and modified Canny edge detector in order to reduce computational cost. Third, the eyelids which are constrained to be within the outer boundary are estimated using the polynomial fitting method. The segmentation method was implemented and tested on iris database set which is captured by hand-held optical sensor device. Experimental results show that the proposed algorithm can separate the iris from the surrounding noises with good speed and accuracy. [ABSTRACT FROM AUTHOR]

We propose an iris recognition system for the identification of persons using support vector machines. Canny's edge detection and the Hough transform are used to find the iris/pupil boundary and a simple thresholding method is employed for eyelash detection. The Gabor wavelet technique is deployed in order to extract the deterministic features in the transformed iris of a person in the form of template. The extracted iris features are fed into a support vector machine (SVM) for classification. Our results indicate that the performance of SVM as a classifier is far better than the performance of a classifier based on the artificial neural network. [ABSTRACT FROM AUTHOR]

In this paper, we present a novel system for ear identification from profile images of the face. The system has two steps. In the first step, the ear is automatically detected from the profile image of the face. In the second step, the ear image is transformed to a force field, then feature points are extracted and the best match is found from a database. We propose a method based on differential geometry to extract ear feature points. We use a transformation of the ear image to make it suitable for extracting the feature points using differential geometry. During recognition, the feature points obtained from a query image are aligned and compared with those in the database using Hausdorff distance. The experimental results show that our method is effective. [ABSTRACT FROM AUTHOR]

In this paper, we firstly present a novel footprint range image segmentation method using the principal curvatures and the principal directions. Utilizing the principal curvatures information, we detect the peak areas as the seeds, and apply region growing to locate the edges of each patch. We apply the edge detection technology to the region growth rules, so the boundary localization is precise. To obtain more stable edge information, a multi-scale fusion approach is proposed to integrate the segmentation results calculated at different fitting sizes. After the segmentation, according to the shape characteristics of footprint, we use superquadric and saddle models to describe shape features of each patch. The experiments results on footprint range images show that the segmented patches and the descriptions represent footprint biometric information effectively and set a reliable basis for the further recognition. [ABSTRACT FROM AUTHOR]

Today, iris recognition, fingerprint recognition, face recognition, voice recognition and other biometric technology are experiencing rapid development. This paper addresses a new biometric technology-the identification and recognition based on point of blood vessel skeleton for ocular fundus. The image for green gray scale of ocular fundus is utilized. The cross point of skeleton shape of blood vessel for ocular fundus using contrast-limited adaptive histogram equalization is extracted at first. After filtering treatment and extracting shape, shape curve of blood vessels is obtained. The cross point of shape for curve matching is later carried out by means of cross point matching. The recognition based on shape for blood vessel of ocular fundus has been demonstrated in this paper to possess high Identification and recognition rate, low rejection recognition rate as well as good universality, exclusiveness and stability. With more and more progress made in extracting technology, the recognition for blood vessel of optic fundus is to become an effective biometric technology. [ABSTRACT FROM AUTHOR]

In this paper, we present a new method for authentication of biometric images. Our method uses an invertible watermark that can also detect malicious manipulations simultaneously. While virtually all watermarking schemes introduce a small amount of non-invertible distortion in original biometric images, our new method is invertible in the sense that, if the data is deemed authentic, distortion due to authentication can be removed if it becomes necessary to obtain the original biometric image. This technique provides cryptographic strength when verifying image integrity because the probability of making an undetectable modification to the image can be directly related to a secure cryptographic element, such as a hash function. Also, if the biometric image is manipulated, the positions of intentional manipulation can be clearly identified. [ABSTRACT FROM AUTHOR]

Biometric Person Authentication such as face, fingerprint, palmprint and signature depends on the quality of image processing. When it needs to be done under a low-resolution image, the accuracy will be impaired. So how to recover the lost information from downsampled images is important for both authentication and preprocessing. Based on Super-Resolution through Neighbor Embedding algorithm and histogram matching, we propose an improved super-resolution approach to choose more reasonable training images. First, the training image are selected by histogram matching. Second, neighbor embedding algorithm is employed to recover the high-resolution image. Experiments in several images show that our improved super-resolution approach is promising for potential applications such as low-resolution mobile phone or CCTV (Closed Circuit Television) image person authentication. [ABSTRACT FROM AUTHOR]

We present a new human identification system based on PCA using geometric features of teeth such as the size and shape of the jaws, size of the teeth and teeth structure. In this paper we try to set forth the foundations of a biometric system for information encrypting of living people using dental features. To create a biometric matching system, a template based on principal component analysis(PCA) is created from dental data collected the plaster figures of teeth which were done at dental hospital, department of oral medicine. Templates of dental images based on PCA representation include the 100 principle components as the features for individual identification. The PCA basis vectors reflects well the features for individual identification in the whole of teeth and the part of teeth. The classification for human identification is generated based on the distance between the whole of teeth and the part of teeth with the nearest neighbor(NN) algorithm. The identification performance in 300 dental image is 97% for the part of teeth missed the right-molar and back teeth, 98.3% for the part of teeth missed the front teeth and 96.6% for the part of teeth missed the left-molar and back-teeth. [ABSTRACT FROM AUTHOR]

This paper presents a novel method of feature-level fusion (FLF) based on kernel principle component analyze (KPCA). The proposed method is applied to fusion of hand biometrics include palmprint, hand shape and knuckleprint, and we name the new feature as "handmetric". For different kind of samples, polynomial kernel is employed to generate the kernel matrixes that indicate the relationship among them. While fusing these kernel matrixes by fusion operators and extracting principle components, the handmetric feature space is established and nonlinear feature-level fusion projection could be implemented. The experimental results testify that the method is efficient for feature fusion, and could keep more identity information for verification. [ABSTRACT FROM AUTHOR]

Hand-based biometrics such as fingerprint and palmprint had been widely accepted because of their convenience and ease in usage without intruding much to one's privacy such as face. The aim of this work is to develop a new point-based algorithm for palmprint feature extraction and perform reliable verification based on the extracted features. This point-based recognition system is next used as part of a bimodal palmprint recognition system combining with a DCT-based (Discrete Cosine Transform) algorithm for identity verification. The performance of the integrated system is evaluated using physical palmprint images. Keywords: Biometrics, Palmprint Recognition, Multimodal Biometrics, and Identity Verification. [ABSTRACT FROM AUTHOR]

For applications like Terrorist Watch Lists and Smart Guns, a false rejection is more critical than a false acceptance. In this paper a new threat model focusing on false rejections is presented, and the "standard" architecture of a biometric system is extended by adding components like crypto, audit logging, power, and environment to increase the analytic power of the threat model. Our threat model gives new insight into false rejection attacks, emphasizing the role of an external attacker. The threat model is intended to be used during the design of a system. [ABSTRACT FROM AUTHOR]

This paper presents a biometric recognition system based on hand geometry. We describe a database specially collected for research purposes, which consists of 50 people and 10 different acquisitions of the right hand. This database can be freely downloaded. In addition, we describe a feature extraction procedure and we obtain experimental results using different classification strategies based on Multi Layer Perceptrons (MLP). We have evaluated identification rates and Detection Cost Function (DCF) values for verification applications. Experimental results reveal up to 100% identification and 0% DCF. [ABSTRACT FROM AUTHOR]

In this paper we have shown that systematic incorporation of decision from various classifiers following a simple decision decomposition rule, gives better decision in comparison to the existing multiple classifier systems. In our method each classifier were graded according to their effectiveness of providing more accurate results. This approach first utilizes the best classifier. If this classifier classifies the test sample into more than one class or fails to classify the test data then the feature next to the best is summoned to finish up the remaining part of the classification. The continuation of this process, along with the judicious selection of classifiers, yields better efficiency in identifying a single class for the test data. The results obtained after the experiments on a set of fingerprint images shows the effectiveness of our proposed classifier. [ABSTRACT FROM AUTHOR]

The purpose of this research was to detect the pattern of player's emotional change during on-line game. By defining data processing technique and analysis method for bio-physiological activity and player's bluffing behavior, the classification of affective attitudes during on-line game was attempted. Bluffing behavior displayed during the game was classified into two dimensions of emotional axis based on prefrontal surface electroencephalographic data. Classified bluffing attitudes were: (1) pleasantness/unpleasantness; and (2) honesty/bluffing. A multilayer-perception neural network was used to classify the player state into four attitude categories. Resulting classifier showed moderate performance with 67.03% pleasantness/unpleasantness classification, and 77.51% for honesty/bluffing. The classifier model developed in this study was integrated to on-line game as a form of ‘emoticon' which displays facial expression of opposing player's emotional state. [ABSTRACT FROM AUTHOR]

We investigate how the acoustic properties of the pinna - i.e. the outer flap of the ear- and the ear canal can be used as a biometric. The acoustic properties can be measured relatively easy with an inexpensive sensor and feature vectors can be derived with little effort. Classification results for three platforms are given (headphone, earphone, mobile phone) using noise as an input signal. Furthermore, preliminary results are given for the mobile phone platform where we use music as an input signal. We achieve equal error rates in the order of 1%-5%, depending on the platform that is used to do the measurement. [ABSTRACT FROM AUTHOR]

Identification of deceased individuals based on dental characteristics is receiving increased attention. Dental radiographic films of an individual are usually composed into a digital image record. In order to achieve high level of automation in postmortem identification, it is necessary to decompose dental image records into their constituent radiographic films, which are in turn segmented to localize dental regions of interest. In this paper we offer an automatic hierarchical treatment to the problem of cropping dental image records into films. Our approach is heavily based on concepts of mathematical morphology and shape analysis. Among the many challenges we face are non-standard assortments of films into records, variability in record digitization as well as randomness of record background both in intensity and texture. We show by experimental evidence that our approach achieves high accuracy and timeliness. [ABSTRACT FROM AUTHOR]

The Fisherface method is a most representative method of the linear discrimination analysis (LDA) technique. However, there persist in the Fisherface method at least two areas of weakness. The first weakness is that it cannot make the achieved discrimination vectors completely satisfy the statistical uncorrelation while costing a minimum of computing time. The second weakness is that not all the discrimination vectors are useful in pattern classification. In this paper, we propose an uncorrelated Fisherface approach (UFA) to improve the Fisherface method in these two areas. Experimental results on different image databases demonstrate that UFA outperforms the Fisherface method and the uncorrelated optimal discrimination vectors (UODV) method. [ABSTRACT FROM AUTHOR]

This paper studies the smartcard based fingerprint encrytion/auth-entication scheme for ATM banking systems. In this scheme, the system authenticates each user by both his/her possession (smartcard) and biometrics (fingerprint). A smartcard is used for the first layer of authentication. Based on the successful pass of the first layer authentication, a subsequent process of the biometric fingerprint authentication proceeds. The proposed scheme is fast and secure. Computer simulations and statistical analyze are presented. [ABSTRACT FROM AUTHOR]

Biometric recognition based on human characteristics for personal identification has attracted great attention. The performance of biometric systems highly depends on the distinctive information in the biometrics. However, identical twins having the closest genetics-based relationship are expected to have maximum similarity between their biometrics. Classifying identical twins is a challenging problem for some automatic biometric systems. In this paper, we summarize the exiting experimental results about identical twins' biometrics including face, iris, fingerprint and voice. Then, we systemically examine identical twins' palmprints. The experimental results show that we can employ low-resolution palmprint images to distinguish identical twins. [ABSTRACT FROM AUTHOR]

Security is a critical component of most computer systems - especially those used in E-commerce activities over the Internet. Global access to information makes security a critical design issue in these systems. Deployment of sophisticated hardware based authentication systems is prohibitive in all but the most sensitive installations. What is required is a reliable, hardware independent and efficient security system. In this paper, we propose an extension to a keystroke dynamics based security system. We provide evidence that completely software based systems based on keystroke input dynamics can be as effective as expensive and cumbersome hardware based systems. Our system is behavioral based that captures the typing patterns of a user and uses that information, in addition to standard login/password security to provide a system that is user-friendly and very effective at detecting imposters. [ABSTRACT FROM AUTHOR]

User authentication based on keystroke dynamics is concerned with accepting or rejecting someone based on the way the person types. A timing vector is composed of the keystroke duration times interleaved with the keystroke interval times. Which times or features to use in a classifier is a classic feature selection problem. Genetic algorithm based wrapper approach does not only solve the problem, but also provides a population of "fit" classifiers which can be used in ensemble. In this paper, we propose to add uniqueness term in the fitness function of genetic algorithm. Preliminary experiments show that the proposed approach performed better than two phase ensemble selection approach and prediction based diversity term approach. [ABSTRACT FROM AUTHOR]

We developed the keystroke biometric system (KBS) using the statistical features of the discrete wavelet transformed keystroke pattern in the frequency domain in addition to those of the original keystroke pattern in the time domain. Only 20 keystroke patterns of user's password typing, where the length of password is no more than 10, are used for building a KBS. The features in the time domain and those in the frequency domain are separately scored by the rules that we developed, and arbitrary given keystroke patterns are classified on the basis of total scores. The results show that our KBS is competitive in comparison with others due to its cheap computational cost, cheap usability cost, and the practically acceptable classification accuracy. Keywords: Keystroke Dynamics, Keystroke Biometric System, Key- stroke Authentication, Discrete Wavelet Transform. [ABSTRACT FROM AUTHOR]

This paper presents a new approach for biometric authentication based on keystroke dynamics through numerical keyboards. The input signal is generated in real time when the user enters with target string. Five features were extracted from this input signal (ASCII key code and four keystroke latencies) and four experiments using samples for genuine and impostor users were performed using two pattern classification technics. The best results were achieved by the HMM (EER=3.6%). This new approach brings security improvements to the process of user authentication, as well as it allows to include biometric authentication in mobile devices, such as cell phones. [ABSTRACT FROM AUTHOR]

In keystroke dynamics-based authentication, novelty detection methods have been used since only the valid user's patterns are available when a classifier is built. After a while, however, impostors' keystroke patterns become also available from failed login attempts. We propose to retrain the novelty detector with the impostor patterns to enhance the performance. In this paper the support vector data description (SVDD) and the one-class learning vector quantization (1-LVQ) are retrained with the impostor patterns. Experiments on 21 keystroke pattern datasets show that the performance improves after retraining and that the one-class learning vector quantization outperforms other widely used novelty detectors. [ABSTRACT FROM AUTHOR]

Biometrics based user authentication involves collecting user's patterns and then using them to determine if a new pattern is similar enough. The quality of the user's patterns is as important as the quality of the classifier. But, the issue has been ignored in the literature since the popular biometrics are mostly trait based such as finger prints and iris so that its pattern quality depends on the quality of the input device involved. However, the quality of the user's patterns of behavior based biometric such as keystroke dynamics can be improved artificially by increasing the peculiarity of the typing style. In this paper, we propose several ways to improve the quality. But, first we define the quality of patterns in terms of two factors: uniqueness and consistency. Finally, the results of a preliminary experiment are presented that support the utility of the proposed methods. [ABSTRACT FROM AUTHOR]

In this paper, we propose a novel method to reconstruct 3D human body pose for gait recognition from monocular image sequences based on top-down learning. Human body pose is represented by a linear combination of prototypes of 2D silhouette images and their corresponding 3D body models in terms of the position of a predetermined set of joints. With a 2D silhouette image, we can estimate optimal coefficients for a linear combination of prototypes of the 2D silhouette images by solving least square minimization. The 3D body model of the input silhouette image is obtained by applying the estimated coefficients to the corresponding 3D body model of prototypes. In the learning stage, the proposed method is hierarchically constructed by classifying the training data into several clusters recursively. Also, in the reconstructing stage, the proposed method hierarchically reconstructs 3D human body pose with a silhouette image. The experimental results show that our method can be efficient and effective to reconstruct 3D human body pose for gait recognition. [ABSTRACT FROM AUTHOR]

Recently, gait recognition for human identification has received substantial attention from biometrics researchers. Compared with other biometrics, it is more difficult to disguise. In addition, gait can be captured in a distance by using low-resolution capturing devices. In this paper, we proposed a new representation for human gait recognition which is called Motion Silhouettes Image (MSI). MSI is a grey-level image which embeds the critical spatio-temporal information. Experiments showed that MSI has a high discriminative power for gait recognition. The recognition rate is around 87% in SOTON dataset by using MSI for recognition. The recognition rate is quite promising. In addition, MSI can also reduce the storage size of the dataset. After using MSI, the storage size of SOTON has reduced to 4.2MB. [ABSTRACT FROM AUTHOR]

In recent years, many gait recognition algorithms have been developed, but most of them depend on a specific view angle. In this paper,we present a new gait recognition scheme based on multi-view gait sequence fusion. An experimental comparison of the fusion of gait sequences at different views is reported. Our experiments show the fusion of gait sequences at different views can consistently achieve better results. The Dempster-Shafer fusion method is found to give a great improvement. On the other hand, we also find that fusion of gait sequences with an angle difference greater than or equal to 90° can achieve better improvement than fusion of those with an acute angle difference. [ABSTRACT FROM AUTHOR]

We have analyzed the vulnerability and threat of the biometric evaluation database and proposed the method to generate a synthetic database from a real database. Our method is characterized by finding nearest neighbor triples or pairs in the feature space of biometric samples, and by crossing over those triples and pairs to generate synthetic samples. The advantages of our method is that we can keep the statistical distribution of the original database, thus, the evaluation result is expected to be the same as original real database. The proposed database, which does not have privacy problem, can be circulated freely among biometric vendors and testers. We have implemented this idea on a face image database using active appearance model. The synthesized image database has the same distance distribution with the original database, which suggests it will deriver the same accuracy with the original one. [ABSTRACT FROM AUTHOR]

For biometric person authentication, evaluations on a biometric system are very essential parts of the entire process. This paper presents the technology evaluations on the TH-FACE recognition system. The main objectives of the evaluations are to 1) test the performance of the TH-FACE recognition system objectively; 2) provide a method to design and organize a database for evaluations; 3) identify the advantage and weakness for the TH-FACE recognition system. Particular description of the test database used in the evaluations is given in this paper. The database contains different subsets which are sorted by different poses, illuminations, ages, accessory, etc. Results and analysis on the entire performances of the TH-FACE recognition system would be also presented. [ABSTRACT FROM AUTHOR]

The current paper demonstrates the role of statistical models in authentication tasks - both in system development and in performance evaluation. We first introduce a model-based face authentication system based on the Fourier domain phase using Gaussian Mixture Models (GMM) which yields verification error rates as low as 0.3% on a face database of 65 individuals with extreme illumination variations. We then present a statistical framework for predicting authentication error rates for future populations in a rigorous way. This is in contrast to most evaluation protocols used today that are based on observational studies and valid only for the databases at hand. Applications establish that our model-based approach has better predictive performance than an existing state-of-the-art authentication technique. [ABSTRACT FROM AUTHOR]

The recognition accuracy of a single biometric authentication system is often much reduced due to the environment, user mode and physiological defects. In this paper, we combine face and iris features for developing a multimode biometric approach, which is able to diminish the drawback of single biometric approach as well as to improve the performance of authentication system. We combine a face database ORL and iris database CASIA to construct a multimodal biometric experimental database with which we validate the proposed approach and evaluate the multimodal biometrics performance. The experimental results reveal the multimodal biometrics verification is much more reliable and precise than single biometric approach. Keywords: Multimodal biometrics, face, iris, wavelet probabilistic neural network. [ABSTRACT FROM AUTHOR]

In this paper we describe a system that continually verifies the presence/participation of a logged-in user. This is done by integrating multimodal passive biometrics in a Bayesian framework that combines both temporal and modality information holistically, rather than sequentially. This allows our system to output the probability that the user is still present even when there is no observation. Our implementation of the continuous verification system is distributed and extensible, so it is easy to plug in additional asynchronous modalities, even when they are remotely generated. Based on real data resulting from our implementation, we find the results to be promising. [ABSTRACT FROM AUTHOR]

Human faces provide demographic information, such as gender and ethnicity. Different modalities of human faces, e.g., range and intensity, provide different cues for gender and ethnicity identifications. In this paper we exploit the range information of human faces for ethnicity identification using a support vector machine. An integration scheme is also proposed for ethnicity and gender identifications by combining the registered range and intensity images. The experiments are conducted on a database containing 1240 facial scans of 376 subjects. It is demonstrated that the range modality provides competitive discriminative power on ethnicity and gender identifications to the intensity modality. For both gender and ethnicity identifications, the proposed integration scheme outperforms each individual modality. [ABSTRACT FROM AUTHOR]

In this paper, an identity verification framework which combines pattern information from the palm-vein and the palm-crease texture is proposed. Main feature of this system is the use of a low cost Near-Infra-Red (NIR) camera instead of the more expensive infra-red thermal camera for palm image capture. Our preliminary experiments show that useful information from palm-vein and palm-crease texture can be effectively extracted for identity verification using a simple setup to contain the camera. Keywords:Biometrics, Multimodal Biometrics, Palm-vein Recognition, Palm-print Recognition and Pattern Classification. [ABSTRACT FROM AUTHOR]

Audio-Visual speaker recognition promises higher performance than any single modal biometric systems. This paper further improves the novel approach based on Dynamic Bayesian Networks (DBNs) to bimodal speaker recognition. In the present paper, we investigate five different topologies of feature-level fusion framework using DBNs. We demonstrate that the performance of multimodal systems can be further improved by modeling the correlation of between the speech features and the face features appropriately. The experiment conducted on a multi-modal database of 54 users indicates promising results, with an absolute improvement of about 7.44% in the best case and 3.13% in the worst case compared with single modal speaker recognition system. [ABSTRACT FROM AUTHOR]

In speaker verification, it is necessary to reduce the influence of different environmental conditions. In this paper, two stages of normalization techniques, feature normalization and score normalization, are examined for decreasing the mismatch between training and testing acoustic conditions. At the first stage, cepstral mean and variance normalization (CMVN) is modified to normalize the cepstral coefficients with the similar segmental parameter statistics. Next, due to score variability between verification trials, Test-dependent zero-score normalization (TZnorm) and Zero-dependent test-score normalization (ZTnorm) are comparatively presented to transform the output scores entirely and make the speaker-independent decision threshold more robust under adverse conditions. Experiments on NIST2002 SRE corpus show that the normalizations with CMVN in feature stage and ZTnorm in score stage achieved 20.3% relative reduction of EER and 18.1% relative reduction of the minimal DCF compared to the baseline system using CMN and zero normalization. [ABSTRACT FROM AUTHOR]

Due to the fact that the biometric characteristics of a person are bound to that person, biometric methods deployed for signer authentication have the potential of improving the binding of electronic signatures to persons. If there is evi dence that a biometric method was used for signer authentication, and if the level of security of this method is suffi ciently high, then the receiver of a signed docu ment can trust that the signature creation was indeed initiated by the legitimate holder of the private signature key. To achieve this goal, an approach to provide evidence of the use of biometric signer authentication has been developed. The approach has been implemented in a prototype electronic signature creation system with fingerprint verification. [ABSTRACT FROM AUTHOR]

Writing forces are important dynamics of online signatures and it is harder to be imitated by forgers than signature shapes. An improved DTW (Dynamic Time Warping) algorithm is put forward to verify online signatures based on writing forces. Compared to the general DTW algorithm, this one deals with the varying consistency of signature point, signing duration and the different weights of writing forces in different direction. The iterative experiment is introduced to decide weights for writing forces in different direction and the classification threshold. A signature database is constructed with F_Tablet and the experiments results are present in the end. [ABSTRACT FROM AUTHOR]

In this paper, we present a method for generating cryptographic keys that can be replaced if the keys are compromised and without requiring a template signature to be stored. The replaceability of keys is accomplished using iterative inner product of Goh-Ngo [1] Biohash method, which has the effect of re-projecting the biometric into another subspace defined by user token. We also utilized a modified Chang et al [2] Multi-state Discretization (MSD) method to translate the inner products into binary bit-strings. Our experiments indicate encouraging result especially for skilled and random forgery whereby the equal error rates are <6.7% and ~0% respectively, indicating that the keys generated are sufficiently distinguishable from impostor keys. [ABSTRACT FROM AUTHOR]

In this paper, two new methods for online signature verification are proposed. The methods adopt the idea of the longest common subsequences (LCSS) algorithm to a kernel function for Support Vector Machines (SVM). The two kernels LCSS-global and LCSS-local offer the possibility to classify time series of different lengths with SVM. The similarity of two time series is determined very accurately since outliers are ignored. Consequently, LCSS-global and LCSS-local are more robust than algorithms based on dynamic time alignment such as Dynamic Time Warping (DTW). The new methods are compared to other kernel-based methods (DTW-kernel, Fisher-kernel, Gauss-kernel). Our experiments show that SVM with LCSS-local and LCSS-global authenticate persons very reliably. [ABSTRACT FROM AUTHOR]

This paper explores the fusion of audio and visual evidences through a multi-level hybrid fusion architecture based on dynamic Bayesian network (DBN), which combines model level and decision level fusion to achieve higher performance. In model level fusion, a new audio-visual correlative model (AVCM) based on DBN is proposed, which describes both the inter-correlations and loose timing synchronicity between the audio and video streams. The experiments on the CMU database and our own homegrown database both demonstrate that the methods can improve the accuracies of audio-visual bimodal speaker identification at all levels of acoustic signal-to-noise-ratios (SNR) from 0dB to 30dB with varying acoustic conditions. [ABSTRACT FROM AUTHOR]

This paper presents an intensity-based iris recognition system. The system exploits local intensity changes of the visible iris textures such as crypts and naevi. The textures are extracted using local histogram equalization and the proposed ‘quotient thresholding' technique. The quotient thresholding partitions iris images in a database such that a ratio between foreground and background of each image is retained. By fixing this ratio, variations of illumination across iris images are compensated, resulting in informative and distinctive blob-like iris textures. An agreement of the two extracted textures is measured by finding spatial correspondences between the textures. The proposed system yields the 0.22 %EER and 100%CRR. The experimental results indicate encouraging and effective iris recognition system, especially when it is used in identification mode. The system is very robust to changes in decision ratio. [ABSTRACT FROM AUTHOR]

As the security of personal information is becoming more important in mobile phones, we apply iris recognition technology to mobile device. Different from conventional iris recognition system used for access control, user puts the mobile phone by hands in this case. So, optical and motion blurring happens, frequently. In addition, most users have tendencies to use the mobile phone in outdoor and sunlight (which includes much amount of IR(Infra-Red) light) may have much effect on the input iris image in spite of the visible light cut filter attached in front of iris camera lens. To overcome such problems, we propose a new method of extracting the accurate iris code based on AGF (Adaptive Gabor Filter). The kernel size, frequency and amplitude of Gabor filter are determined by the amount of blurring and sunlight in input image, adaptively. Experimental results show that the EER by our propose method is 0.14 %. [ABSTRACT FROM AUTHOR]

Iris recognition is one of the most reliable methods for personal identification. However, not all the iris images obtained from the device are of high quality and suitable for recognition. In this paper, a novel approach for iris image quality assessment is proposed to select clear images in the image sequence. The proposed algorithm uses three distinctive features to distinguish three kinds of poor quality images, i.e. defocus, motion blur and occlusion. Experimental results demonstrate the effectiveness of the algorithm. Clear iris images selected by our method are essential to subsequent iris recognition. [ABSTRACT FROM AUTHOR]

A practical iris identification application system faces different types of bad iris images resulted from many reasons. Because previous image quality evaluation methods estimate an iris image whether bad or else by the resolution and the definition of the iris part, they just can deal with few types among them. For saving the time occupied by the localization in images real-time estimation, improving friendly interaction of an iris identification system, decreasing the localization failure on account of importing the bad-image, this paper proposes a method of real-time pre-estimation using the compound BP neural network. Multiple independent BP neural networks are used to extract both the overall contour feature and the local of an iris image and to calculate the pre-estimation output by different training weights. The experimental result is shown that the method can detects most types of the bad-image with comparatively low error rate and the pre-estimation network has fairly large throughput. It should satisfy the pre-estimation requirement of a real-time iris identification system. [ABSTRACT FROM AUTHOR]

The method of Wavelet Packets Decomposition (WPD) originating from wavelet transform is more accurate in signal analysis, with the predominance of analyzing high-frequency information. Combined with the trait of WPD, an algorithm for iris recognition is presented in this paper. Firstly, iris image is divided into several windows, and WPD is done to them. At the same time, some of the subband images from each window are selected, which contain most information of iris image. Secondly, the farther feature extraction and compression are applied to these subband images by way of Singular Value Decomposition (SVD), and iris recognition features are obtained. Finally, Weighted Euclidean Distance (WED) classifier is utilized in recognition. Experimental results on CASIA (Chinese Academy of Sciences, Institute of Automation) iris image database show the method is valid in iris recognition. [ABSTRACT FROM AUTHOR]

Iris location is a crucial step in iris recognition. Taking into consideration the fact that interior of the pupil, there would have some lighter spots because of reflection, this paper improves the commonly used coarse location method. It utilizes the gray scale histogram of the iris graphics, first computes the binary threshold, averaging the center of chords to coarsely estimate the center and radius of the pupil, and then finely locates it using the algorithm of circle detection in the binary graphic. This method could reduce the error of locating within the pupil. After that, this paper combines Canny edge detector and Hough voting mechanism to locate the outer boundary. Finally, a statistical method is exploited to exclude eyelash and eyelid areas. Experiments have shown the applicability and efficiency of this algorithm. Keywords: Iris Location, Circle Detection, Canny Edge Detection, Hough Voting Mechanism. [ABSTRACT FROM AUTHOR]

Intra-class variation is fundamental to the FNMR performance of iris recognition systems. In this paper, we perform a systematic study of modeling intra-class variation for nonideal iris images captured under less-controlled environments. We present global geometric calibration techniques for compensating distortion associated with off-angle acquisition and local geometric calibration techniques for compensating distortion due to inaccurate segmentation or pupil dilation. Geometric calibration facilitates both the localization and recognition of iris and more importantly, it offers a new approach of trading FNMR with FMR. We use experimental results to demonstrate the effectiveness of the proposed calibration techniques on both ideal and non-ideal iris databases. [ABSTRACT FROM AUTHOR]