This paper presents an R&D framework used by the National Institute of Standards and Technology (NIST) for biometric technology testing and evaluation. The focus of this paper is on fingerprint-based verification and identification. Since 9/11 the NIST Image Group has been mandated by Congress to run a program for biometric technology assessment and biometric systems certification. Four essential areas of activity are discussed: 1) developing test datasets; 2) conducting performance assessment; 3) technology development; and 4) standards participation. A description of activities and accomplishments are provided for each of these areas. In the process, methods of performance testing are described and results from specific biometric technology evaluations are presented. This framework is anticipated to have broad applicability to other technology and application domains

Over the last five years or so, neural network (NN)-based approaches have been steadily gaining performance and popularity for a wide range of optical character recognition (OCR) problems, from isolated digit recognition to handprint recognition. We present an NN classification scheme based on an enhanced multilayer perceptron (MLP) and describe an end-to-end system for form-based handprint OCR applications designed by the National Institute of Standards and Technology (NIST) Visual Image Processing Group. The enhancements to the MLP are based on (i) neuron activations functions that reduce the occurrences of singular Jacobians; (ii) successive regularization to constrain the volume of the weight space; and (iii) Boltzmann pruning to constrain the dimension of the weight space. Performance characterization studies of NN systems evaluated at the first OCR systems conference and the NIST form-based handprint recognition system are also summarized.

To successfully apply character recognition technology most of the forms currently hand-keyed will need to be redesigned. This paper presents results from a comprehensive study of three versions of a redesigned tax form. Analyses show that using separately spaced character boxes provides superior machine readability over fields containing combs and adjoining character boxes. It is shown that character boxes containing two vertically stacked ovals cause writers much more difficulty. Analyses provide proof that writer idiosyncratic responses on forms are the major source of errors, and proper form design can reduce these errors.

This paper introduces scoring methods developed to automatically assess the performance of document recognition systems, specifically, to evaluate the spatial correspondence of zones produced by a document segmentor. Two different approaches are discussed. The first approach (based on zone overlap and nearest-neighbors) is better applied to merged zones, whereas the second approach (based on zone alignments) is better applied to nested zones (such as those found in tables and graphs). Definitions of coverage and efficiency error are presented, and scoring results on real system output is provided that validates the usefulness of these methods to compare different document recognition algorithms. Currently, no standard testing procedures exist for measuring and comparing algorithms within a complex document recognition system. Scoring methods, like the ones introduced in this paper, serve as design and validations tools, expediting the development and deployment of document analysis technology for system developers and end users.

Two neural-network-based methods are combined to develop font-independent character recognition on a distributed array processor. Feature localization and noise reduction are achieved using least-squares optimized Gabor filtering. The filtered images are then presented to an ART-1-based learning algorithm which produces self-organizing sets of neural network weights used for character recognition. Implementation of these algorithms on a highly parallel computer with 1024 processors allows high-speed character recognition to be achieved in 8 ms/image with greater than 99% accuracy on machine print and 80% accuracy on unconstrained hand-printed characters