This paper presents a blind algorithm for the automatic detection of isolated astrophysical pulses. The detection algorithm is applied to spectrograms (also known as “filter bank data” or “the (t,f) plane”). The detection algorithm comprises a sequence of three steps: (1) a Radon transform is applied to the spectrogram, (2) a Fourier transform is applied to each projection parametrized by an angle, and the total power in each projection is calculated, and (3) the total power of all projections above 90° is compared to the total power of all projections below 90° and a decision in favor of an astrophysical pulse present or absent is made. Once a pulse is detected, its Dispersion Measure (DM) is estimated by fitting an analytically developed expression for a transformed spectrogram containing a pulse, with varying value of DM, to the actual data. The performance of the proposed algorithm is numerically analyzed.

This paper investigates the problem of linear spatial collaboration for distributed estimation in wireless sensor networks. In this context, the sensors share their local noisy (and potentially spatially correlated) observations with each other through error-free, low cost links based on a pattern defined by an adjacency matrix. Each sensor connected to a central entity, known as the fusion center (FC), forms a linear combination of the observations to which it has access and sends the resulting signal to the FC through an orthogonal fading channel. The FC combines these received signals to find the best linear unbiased estimator of the vector of unknown signals observed by individual sensors. The main novelty of this paper is the derivation of an optimal power-allocation scheme in which the coefficients used to form linear combinations of noisy observations at the sensors connected to the FC are optimized. Through this optimization, the total estimation distortion at the FC is minimized, given a constraint on the maximum cumulative transmit power in the entire network. Numerical results show that even with a moderate connectivity across the network, spatial collaboration among sensors significantly reduces the estimation distortion at the FC.

Partial face recognition has been a problem of interest for more than a decade. Most of previous publications on partial face recognition assume intra spectral matching. Matching Short Wave Infrared (SWIR), Middle Wave Infrared (MWIR) or Near Infrared (NIR) images of partial face to a gallery of color images is a much more challenging task. The photometric properties of images in these four spectral bands are highly distinct. Because of the limited space - and also sufficient interest to this biometric - in this paper we present results of cross spectral matching applied to periocular regions. Equipped with a well developed automatic recognition algorithm for heterogeneous face, we demonstrate that the algorithm can be tuned and applied to periocular regions for a positive cross spectral matching of SWIR, MWIR and NIR periocular regions to visible periocular regions at short (1.5 m) and long (50 and 106 m) standoff distances. Our numerical analysis demonstrates the results of the matching. To the best of our knowledge, the performance evaluation presented in this paper is the first of its kind.

Three practical methods to improve performance of a single biometric matcher based on vectors of quality measures associated with biometric data are described. The first two methods adaptively select probe biometric data and matching scores based on predicted values of Quality of Sample (QS) index (defined here as d-prime) and Confidence in matching Scores (CS), respectively. The third method, Quality Sample and Template features (QST), treats quality measures as weak but useful features for discriminating between genuine and imposter matching scores. The unifying theme for the three methods consists in learning a nonlinear mapping between vectors of quality measures and QS, CS, and QST for each of the three methods, respectively. For the first method, learning requires a small set of input data in the form of a vector of quality metrics per each biometric image and the output data in the form of QS estimated per image. For the second method, learning requires a small set of input data in the form of two vectors of quality metrics per each matching pair and the output data in the form of CS estimated per matching score. For the third method, learning requires a small set of input data in the form of biometric feature vector (template) concatenated with a vector of quality metrics and a set of output data in the form of matching labels. The proposed methodology is generic and is suitable for any biometric modality and for any choice of a nonlinear mapping between vectors of quality measures and QS, CS, and QST. The experimental results (obtained by means of neural nets) show significant performance improvements for all three methods when applied to iris biometrics.

Short wave infrared (SWIR) is an emerging imaging modality in surveillance applications. It is able to capture clear long range images of a subject in harsh atmospheric conditions and at night time. However, matching SWIR images against a gallery of color images is a very challenging task. The photometric properties of images in these two spectral bands are highly distinct. This work presents a novel cross-spectral face recognition scheme that encodes images filtered with a bank of Gabor filters followed by three local operators: Simplified Weber Local Descriptor, Local Binary Pattern, and Generalized Local Binary Pattern. Both magnitude and phase of filtered images are encoded. Matching encoded face images is performed by using a symmetric I-divergence. We quantify the verification and identification performance of the cross-spectral matcher on two multispectral face datasets. In the first dataset (PRE-TINDERS), both SWIR and visible gallery images are captured at a close distance (about 2 meters). In the second dataset (TINDERS), the probe SWIR images are collected at longer ranges (50 and 106 meters). The results on PRE-TINDERS dataset form a baseline for matching long range data. We also demonstrate the capability of the proposed approach by comparing its performance with the performance of Faceit G8, a commercial face recognition engine distributed by L1. The results show that the designed method outperforms Faceit G8 in terms of verification and identification rates on both datasets.

Three methods to improve the performance of biometric matchers based on vectors of quality measures associated with biometric samples are described. The first two methods select samples and matching scores based on predicted values of Quality of Sample (QS) index (defined here as d-prime) and Confidence in matching Scores (CS), respectively. The third method treats quality measures as weak but useful features for discrimination between genuine and imposter matching scores. The unifying theme for the three methods consists of a nonlinear mapping between quality measures and the predicted values of QS, CS, and combined quality measures and matching scores, respectively. The proposed methodology is generic and is suitable for any biometric modality. The experimental results reported show significant performance improvements for all the three methods when applied to iris biometrics.

Iris recognition, the ability to recognize and distinguish individuals by their iris pattern, is one of the most reliable biometrics in terms of recognition and identification performance. However, the performance of these systems is affected by poor-quality imaging. In this paper, we extend iris quality assessment research by analyzing the effect of various quality factors such as defocus blur, off-angle, occlusion/specular reflection, lighting, and iris resolution on the performance of a traditional iris recognition system. We further design a fully automated iris image quality evaluation block that estimates defocus blur, motion blur, off-angle, occlusion, lighting, specular reflection, and pixel counts. First, each factor is estimated individually, and then, the second step fuses the estimated factors by using a Dempster-Shafer theory approach to evidential reasoning. The designed block is evaluated on three data sets: Institute of Automation, Chinese Academy of Sciences (CASIA) 3.0 interval subset, West Virginia University (WVU) non-ideal iris, and Iris Challenge Evaluation (ICE) 1.0 dataset made available by National Institute for Standards and Technology (NIST). Considerable improvement in recognition performance is demonstrated when removing poor-quality images selected by our quality metric. The upper bound on computational complexity required to evaluate the quality of a single image is O(n2log n).

Iris biometric is one of the most reliable biometrics with respect to performance. However, this reliability is a function of the ideality of the data. One of the most important steps in processing nonideal data is reliable and precise segmentation of the iris pattern from remaining background. In this paper, a segmentation methodology that aims at compensating various nonidealities contained in iris images during segmentation is proposed. The virtue of this methodology lies in its capability to reliably segment nonideal imagery that is simultaneously affected with such factors as specular reflection, blur, lighting variation, occlusion, and off-angle images. We demonstrate the robustness of our segmentation methodology by evaluating ideal and nonideal data sets, namely, the Chinese Academy of Sciences iris data version 3 interval subdirectory, the iris challenge evaluation data, the West Virginia University (WVU) data, and the WVU off-angle data. Furthermore, we compare our performance to that of our implementation of Camus and Wildes's algorithm and Masek's algorithm. We demonstrate considerable improvement in segmentation performance over the formerly mentioned algorithms.

The ability of practical recognition systems to recognize a large number of objects is constrained by a variety of factors that include choice of a feature extraction technique, quality of images, complexity and variability of underlying objects and of collected data. Given a feature extraction technique generating templates of objects from data and a resolution of the original images, the remaining factors can be attributed to distortions due to a recognition channel. We define the recognition channel as the environment that transforms reference templates of objects in a database into templates submitted for recognition. If templates in an object database are generated to be statistically independent and the noise in a query template is statistically independent of templates in the database, then the abilities of the recognition channel to recognize a large number of object classes can be characterized by a number called recognition capacity. In this paper, we evaluate the empirical recognition capacity of PCA-based object recognition systems. The encoded data (templates) and the additive noise in query templates are modeled to be Gaussian distributed with zero mean and estimated variances. We analyze both the case of a single encoded image and the case of encoded correlated multiple images. For this case, we propose a model that is orientation and elevation angle (pose) dependent. The fit of proposed models is judged using statistical goodness of fit tests. We define recognition rate as the ratio R=log(M)/n, where M is the number of objects to recognize and n is the length of PCA templates. The empirical capacity of PCA-based recognition systems is numerically evaluated. The empirical random coding exponent is also numerically evaluated and plotted as a function of the recognition rate. With these results, given a value of the recognition capacity and the length of templates (assume large), we can predict the number of distinct object classes that can be stored in an object library and be identified with probability of error close to zero.

The ability of practical biometric systems to recognize a large number of subjects is constrained by a variety of factors that include a choice of a source encoding technique, quality of images, complexity and variability of underlying patterns and of collected data. Given a source encoding technique, the remaining factors can be attributed to distortions due to a biometric recognition channel. In this work, we define empirical mutual information and recognition rate and evaluate empirical recognition capacity of biometric systems under the constraint of two global encoding techniques: principal component analysis (PCA) and independent component analysis (ICA). The empirical capacity of biometric systems is numerically evaluated as a point of intersection of the empirical mutual information rate curve plotted as a function of the recognition rate and the diagonal line bisecting the first quadrant. The developed methodology is applied to find the empirical capacity of different recognition channels formed during acquisition of different iris and face databases.

Performance of biometric-based recognition systems depends on various factors: database quality, image preprocessing, encoding techniques, etc. Given a biometric database and a selected encoding method, the recognition capability of a system is limited by the relationship between the number of classes that the recognition system can encode and the length of encoded data describing the template at a specific level of distortion. In this paper, we evaluate empirical recognition capacity of biometric systems under the constraint of two global encoding techniques: principal component analysis (PCA) and independent component analysis (ICA). The developed methodology is applied to predict the capacity of different recognition channels formed during the acquisition of different iris and face databases. The proposed approach relies on data modeling and involves classical detection and information theories. The major contribution is in providing a guideline on how to evaluate capabilities of large-scale biometric recognition systems that are based on PCA and ICA encoding. Recognition capacity can also be promoted as a global quality measure of biometric databases.

The popularity of iris biometric has grown considerably over the past two to three years. It has resulted in the development of a large number of new iris encoding and processing algorithms. Since there are no publicly available large-scale and even medium-size data bases, neither of the newly designed algorithms has undergone extensive testing. The designers claim exclusively high recognition performance when the algorithms are tested on a small amount of data. In a large-scale setting, systems are yet to be tested. Since the issues of security and privacy slow down the speed of collecting and publishing iris data, an optional solution to the problem of algorithm testing is to synthetically generate a large scale data base of iris images. In this work, we describe a model-based method to generate iris images and evaluate the performance of synthetic irises by using a traditional Gabor filter-based iris recognition system. A comprehensive comparison of synthetic and real data is performed at three levels of processing: 1) image level, 2) texture level, and 3) decision level. A sensitivity analysis is performed to conclude on the importance of various parameters involved in generating iris images

The popularity of the iris biometric has grown considerably over the past two to three years. Most research has been focused on the development of new iris processing and recognition algorithms for frontal view iris images. However, a few challenging directions in iris research have been identified, including processing of a nonideal iris and iris at a distance. In this paper, we describe two nonideal iris recognition systems and analyze their performance. The word "nonideal" is used in the sense of compensating for off-angle occluded iris images. The system is designed to process nonideal iris images in two steps: 1) compensation for off-angle gaze direction and 2) processing and encoding of the rotated iris image. Two approaches are presented to account for angular variations in the iris images. In the first approach, we use Daugman's integrodifferential operator as an objective function to estimate the gaze direction. After the angle is estimated, the off-angle iris image undergoes geometric transformations involving the estimated angle and is further processed as if it were a frontal view image. The encoding technique developed for a frontal image is based on the application of the global independent component analysis. The second approach uses an angular deformation calibration model. The angular deformations are modeled, and calibration parameters are calculated. The proposed method consists of a closed-form solution, followed by an iterative optimization procedure. The images are projected on the plane closest to the base calibrated plane. Biorthogonal wavelets are used for encoding to perform iris recognition. We use a special dataset of the off-angle iris images to quantify the performance of the designed systems. A series of receiver operating characteristics demonstrate various effects on the performance of the nonideal-iris-based recognition system.

A new iterative method for tomographic image reconstruction in the presence of high-density objects of unknown shape and position is described. Application of traditional reconstruction methods in this case result in reconstruction of highly degraded images. We state the problem as a deblurring problem and design an iterative algorithm that jointly reconstructs background attenuation function and the shape of unknown object. B-spline functions are used to parameterize the object boundary. Regularization term is used to preserve against unwanted loops in the shape. We have performed a number of experiments to demonstrate potential of our algorithm.

In the absence of real data for extensive testing of newly designed large-scale biometrics recognition systems a number of solutions are possible including use of resampling methods, generation of synthetic data having properties similar to real data of interest, or use of analytical tools to predict the performance. Each of the methods has its own limitations. In this work, we focus on iris biometric. We briefly describe a model based approach to synthesize iris images and focus on performance comparison for synthesized and real iris images. Iris image processing assumes a traditional Gabor filter based encoding approach. Comparison of synthetic and real data is performed at three levels of processing: (1) image level, (2) texture level, and (3) decision level. The results indicate that in most cases the performance of synthesized iris images is comparable to the performance of the real iris images.

Practical iris-based identification systems are easily accessible for data collection at the matching score level. In a typical setting, a video camera is used to collect a single frontal view image of good quality. The image is then preprocessed, encoded, and compared with all entries in the biometric database resulting in a single highest matching score. In this paper, we assume that multiple scans from the same iris are available and design the decision rules based on this assumption. We consider the cases where vectors of matching scores may be described by a Gaussian model with dependent components under both genuine and imposter hypotheses. Two test statistics: the plug-in loglikelihood ratio and the average Hamming distance are designed. We further analyze the performance of filter-based iris recognition systems. The model fit is verified using the Shapiro-Wilk test for normality. We show that the loglikelihood ratio with well-estimated maximum-likelihood parameters in it often outperforms the average Hamming distance statistic. The problem of identification with M iris classes is further stated as an (M+1)ary hypothesis testing problem. We use empirical approach, Chernoff bound, and Large Deviations approach to predict the performance of the iris-based identification system. The bound on the probability of error is evaluated as a function of the number of classes and the number of iris scans per class.

We analyze the performance of an iris based recognition system. We consider a practical setting where matching scores are accessible for collecting data. We assume that multiple scans from the same iris are available and design the decision rules based on this assumption. We show that vectors of matching scores are described by a Gaussian model with dependent components both under the genuine and imposter hypotheses. Two test statistics: the average Hamming distance and the log-likelihood ratio are designed. We show that the log-likelihood ratio with well estimated maximum likelihood parameters in it outperforms the first test statistic. We further use an empirical approach, Chernoff bound, and large deviations approach to predict the performance of the recognition system.

We describe and analyze the performance of a non-ideal iris recognition system. The system is designed to process non-ideal iris images in two steps: (i) estimation of the gaze direction and (ii) processing and encoding of the rotated iris image. We use two objective functions to estimate the gaze direction: Hamming distance and Daugman's integro-differential operator and determine an estimated angle by picking the value that optimizes the selected objective function. After the angle is estimated, the off-angle iris image undergoes geometric transformations involving the estimated angle and is further processed as if it were a frontal view image. The encoding technique developed in this work is based on application of the global independent component analysis (ICA) to masked iris images. We use two datasets: CASIA dataset and a special dataset of off-angle iris images collected at WVU to verify the performance of the encoding technique and angle estimator, respectively. A series of receiver operating characteristics (ROCs) demonstrates various effects on the performance of the non-ideal iris based recognition system implementing the global ICA encoding.

We consider the estimation of an unknown arbitrary 2D object shape from sparse noisy samples of its Fourier transform. The estimate of the closed boundary curve is parametrized by normalized Fourier descriptors (FDs). We use Rissanen's (1998) MDL criterion. to regularize this ill-posed non-linear inverse problem and determine an optimum tradeoff between approximation and estimation errors by picking an optimum order for the FD parametrization. The performance of the proposed estimator is quantified in terms of the area discrepancy between the true and estimated object. Numerical results demonstrate the effectiveness of the proposed approach.

In physical signature authentication schemes, a candidate signature presented for authentication is compared to a previous signature measurement. The uniqueness of a signature is a result of a random physical process underlying the signature formation. We model signatures as realizations of random processes with known statistics. Rate functions for the false alarm and miss probabilities are determined. The case where the candidate signature and previous signature measurement are zero-mean, stationary, and Gaussian is covered in detail, yielding quantitative measures of performance for system design. The binary case is also discussed.

Often recognition systems must be designed with a relatively small amount of training data. Plug-in test statistics suffer from large estimation errors, often causing the performance to degrade as the measurement vector dimension increases. Choosing a better test statistic or applying a method of dimensionality reduction are two possible solutions to this problem. In this paper, we consider a recognition problem where the data for each population are assumed to have the same parametric distribution but differ in their unknown parameters. The collected vectors of data as well as their components are assumed to be independent. The system is designed to implement a plug-in log-likelihood ratio test with maximum-likelihood (ML) estimates of the unknown parameters instead of the true parameters. Because a small amount of data is available to estimate the parameters, the performance of such a system is strongly degraded relative to the performance with known parameters. To improve the performance of the system we define a thresholding function that, when incorporated into the plug-in log-likelihood ratio, significantly decreases the probability of error for binary and multiple hypothesis testing problems for the exponential class of populations. We analyze the modified test statistic and present the results of Monte Carlo simulation. Special attention is paid to the complex Gaussian model with zero mean and unknown variances.