Showing papers on "Signature recognition published in 2004"

Cancelable biometric filters for face recognition

Abstract: In this paper, we address the issue of producing cancelable biometric templates; a necessary feature in the deployment of any biometric authentication system. We propose a novel scheme that encrypts the training images used to synthesize the single minimum average correlation energy filter for biometric authentication. We show theoretically that convolving the training images with any random convolution kernel prior to building the biometric filter does not change the resulting correlation output peak-to-sidelobe ratios, thus preserving the authentication performance. However, different templates can be obtained from the same biometric by varying the convolution kernels thus enabling the cancelability of the templates. We evaluate the proposed method using the illumination subset of the CMU pose, illumination, and expressions (PIE) face dataset. Our proposed method is very interesting from a pattern recognition theory point of view, as we are able to 'encrypt' the data and perform recognition in the encrypted domain that performs as well as the unencrypted case, regardless of the encryption kernel used; we show analytically that the recognition performance remains invariant to the proposed encryption scheme, while retaining the desired shift-invariance property of correlation filters.

Biometric Authentication: A Machine Learning Approach

Abstract: A breakthrough approach to improving biometrics performance Constructing robust information processing systems for face and voice recognition Supporting high-performance data fusion in multimodal systems Algorithms, implementation techniques, and application examples Machine learning: driving significant improvements in biometric performance As they improve, biometric authentication systems are becoming increasingly indispensable for protecting life and property. This book introduces powerful machine learning techniques that significantly improve biometric performance in a broad spectrum of application domains. Three leading researchers bridge the gap between research, design, and deployment, introducing key algorithms as well as practical implementation techniques. They demonstrate how to construct robust information processing systems for biometric authentication in both face and voice recognition systems, and to support data fusion in multimodal systems. Coverage includes: How machine learning approaches differ from conventional template matching Theoretical pillars of machine learning for complex pattern recognition and classification Expectation-maximization (EM) algorithms and support vector machines (SVM) Multi-layer learning models and back-propagation (BP) algorithms Probabilistic decision-based neural networks (PDNNs) for face biometrics Flexible structural frameworks for incorporating machine learning subsystems in biometric applications Hierarchical mixture of experts and inter-class learning strategies based on class-based modular networks Multi-cue data fusion techniques that integrate face and voice recognition Application case studies

Images can be regenerated from quantized biometric match score data

Abstract: We address the possibility of regenerating sample images from stored biometric data, specifically from automatic face recognition algorithms. Such algorithms calculate a match score from comparison of a newly acquired image of a person to a template calculated from previously captured images. Although several vendors of biometric algorithms claim that an image of a person cannot be regenerated from the template, it has been shown that, in general, such regeneration can be performed with a "hill climbing attack". In order to defend against this attack, it is recommended that biometric algorithms emit only quantized match scores. In this paper, we show that it is still possible to regenerate biometric images even if this recommendation is implemented. Each iteration of the algorithm is applied to a quadrant of the sample image. Before each calculation, noise is added to the image in the opposite quadrant, in order to force the match score to a value just below the quantization threshold, providing useful information. Results show this algorithm successfully regenerates images which compare at high match scores for reasonable values of the quantization level. We conclude that the quantization of match score values does not, by itself, protect against the regeneration of images from stored biometric data.

Adaptation in statistical pattern recognition using tangent vectors

Abstract: We integrate the tangent method into a statistical framework for classification analytically and practically. The resulting consistent framework for adaptation allows us to efficiently estimate the tangent vectors representing the variability. The framework improves classification results on two real-world pattern recognition tasks from the domains handwritten character recognition and automatic speech recognition.

Comparison of elastic matching algorithms for online Tamil handwritten character recognition

Abstract: We present a comparison of elastic matching schemes for writer dependent on-line handwriting recognition of isolated Tamil characters. Three different features are considered namely, preprocessed x-y coordinates, quantized slope values, and dominant point coordinates. Seven schemes based on these three features and dynamic time warping distance measure are compared with respect to recognition accuracy, recognition speed, and number of training templates. Along with these results, possible grouping strategies and error analysis is also presented in brief.

An off-line signature verification method based on the questioned document expert's approach and a neural network classifier

Abstract: In an off-line signature verification method based on personal models, an important issue is the number of genuine samples required to train the writer's model. In a real application, we are usually quite limited in the number of samples we can use for training [Cha, S., 2001, Baltzakis, H. et al., 2001, Yingyong, Q. et al., 1994]. Classifiers like the neural network [Baltzakis, H. et al., 2001], the hidden Markov model [Justino, E.J.R. et al., 2001] and the support vector machine [Justino, E.J.R. et al., 2003] need a substantial number of samples to produce a robust model in the training phase. This paper reports on a global method based on only two classes of models, the genuine signature and the forgery. The main objective of this method is to reduce the number of signature samples required by each writer in the training phase. For this purpose, a set of graphometric features and a neural network (NN) classifier are used.

Color and texture features for person recognition

Abstract: The need for automatic visual surveillance is increasing and the research on person recognition systems is more and more supported. As many biometric recognition methods, e.g. face recognition, are based on quite high camera resolutions which are not available in many situations, we examine features as well as classifier techniques for full body recognition. We present our experiments with color and texture features in the application of full body person recognition. On a database of 53 individuals we tested approved features for object recognition as well as MPEG7 color and texture descriptors on a person recognition task. For comparison, we used an RBF network classifier as well as a nearest-neighbor classifier. Our experiments showed that color as well as texture information is important for a person recognition system. Additionally, a combination of these two kind of features results in a performance improvement.

Combination of hidden Markov models with dynamic time warping for speech recognition

Abstract: We combine hidden Markov models of various topologies and nearest neighbor classification techniques in an exponential modeling framework with a model selection algorithm to obtain significant error rate reductions on an isolated word digit recognition task. This work is a preliminary investigation of large scale modeling techniques to be applied to large vocabulary continuous speech recognition.

Methods and systems for slant compensation in handwriting and signature recognition

Abstract: Various computer-implemented methods are provided. One method for signature recognition includes identifying a vertical element of a signature and determining a tilt angle of the vertical element. Tilt angle is defined by a line that is approximately parallel to the vertical element. In addition, the method includes determining a tilt factor. Tilt factor is defined as a horizontal offset between the tilt angle and a reference tilt as a function of distance from a signature baseline. The method further includes altering the signature using the tilt factor and comparing the altered signature to known signature(s) to determine if the altered signature matches one of the known signature(s). Another method includes altering the signature using a predetermined tilt factor and comparing the altered signature to known signature(s). If the altered signature does not match one of the known signature(s), these steps may be repeated with different predetermined tilt factors.

From Off-line to On-line Handwriting Recognition

Abstract: On-line handwriting includes more information on the time order of the writing signal and on the dynamics of the writing process than off-line handwriting. Therefore, on-line recognition systems achieve higher recognition rates. This can be concluded from results reported in the literature, and has been demonstrated empirically as part of this work. We propose a new approach for recovering the time order of the off-line writing signal. Starting from an over-segmentation of the off-line handwriting into regular and singular parts, the time ordering of these parts and recognition of the word are performed simultaneously. This approach, termed “ OrdRec”, is based on a graph description of the handwriting signal and a recognition process using Hidden Markov Models (HMM). A complete omni-scriptor isolated word recognition system has been developed. Using a dynamic lexicon and models for upper and lower case characters, our system can process binary and gray value word images of any writing style (script, cursive, or mixed). Using a dual handwriting data base which features both the on-line and the off-line signal for each of the 30 000 words written by about 700 scriptors, we have shown experimentally that such an off-line recognition system, using the recovered time order information, can achieve recognition performances close to those of an on-line recognition system.

Biometric authentication using online signatures

Abstract: We overview biometric authentication and present a system for on-line signature verification, approaching the problem as a two-class pattern recognition problem. During enrollment, reference signatures are collected from each registered user and cross aligned to extract statistics about that user’s signature. A test signature’s authenticity is established by first aligning it with each reference signature for the claimed user. The signature is then classified as genuine or forgery, according to the alignment scores which are normalized by reference statistics, using standard pattern classification techniques. We experimented with the Bayes classifier on the original data, as well as a linear classifier used in conjunction with Principal Component Analysis (PCA). The classifier using PCA resulted in a 1.4% error rate for a data set of 94 people and 495 signatures (genuine signatures and skilled forgeries).

Method and system for signature recognition biometrics on a smartcard

Abstract: The present invention discloses a system and methods for biometric security using signature recognition biometrics in a smartcard-reader system. The biometric security system also includes a signature scan sensor that detects biometric samples and a device for verifying biometric samples. In one embodiment, the biometric security system includes a smartcard configured with a signature scan sensor. In another embodiment, the system includes a reader configured with a signature scan sensor. In yet another embodiment, the present invention discloses methods for proffering and processing signature samples to facilitate authorization of transactions.

The repeatability of signatures

Abstract: Signatures are the most widely used form of legally binding identification and authentication. The repeatability of a person's signature underpins its recognition and hence usefulness in everyday authentication situations. This study aims to assess the stability of a set of common features used for analysing signatures both within a single capture session and over time (multiple sessions). Secondly, the physical characteristics of signatures which result in the most repeatable performance for each feature are also analyzed. These results have implications for biometric signature verification systems and the document forensic field in that it gives an indication as to the stability of features leading potentially to improved performance and the types of features that should be analyzed given particular characteristics of the signature under investigation.

Off-line handwriting recognition using various hybrid modeling techniques and character n-grams

Abstract: In this paper a system for on-line cursive handwriting recognition is described. The system is based on Hidden Markov Models (HMMs) using discrete and hybrid modeling techniques. Here, we focus on two aspects of the recognition system. First, we present different hybrid modeling techniques, whereas one depends on an information theory-based neural network (MMI-criterion) used as a vector quantizer and the other uses a neural net for estimating the a posteriori probabilities to replace the codebook of a tied-mixture HMM system. This is the first paper where we present this novel approach -called tied posteriors- for handwriting recognition. Second, we demonstrate the usage of a language model, that consists of character n-grams, as an alternative to the recognition with a large dictionary of German words. Our resulting system for character recognition yields significantly better recognition results using an unlimited vocabulary.

Exploiting general knowledge in user-dependent fusion strategies for multimodal biometric verification

Abstract: A novel strategy for combining general and user-dependent knowledge in a multimodal biometric verification system is presented. It is based on SVM classifiers and trade-off coefficients introduced in the standard SVM training problem. Experiments are reported on a bimodal biometric system based on fingerprint and on-line signature traits. A comparison between three fusion strategies, namely user-independent, user-dependent and the proposed adapted user-dependent, is carried out. As a result, the suggested approach outperforms the former ones. In particular, a highly remarkable relative improvement of 68% in the EER with respect to the user-independent approach is achieved. The severe and very common problem of training data scarcity in the user-dependent strategy is also relaxed by the proposed scheme, resulting in a relative improvement of 40% in the EER compared to the raw user-dependent strategy.

A stroke based algorithm for dynamic signature verification

Abstract: Dynamic signature verification (DSV) uses the behavioral biometrics of a hand-written signature to confirm the identity of a computer user. This paper presents a novel stroke-based algorithm for DSV. An algorithm is developed to convert sample signatures to a template by considering their spatial and time domain characteristics, and by extracting features in terms of individual strokes. Individual strokes are identified by finding the points where there is a: 1) decrease in pen tip pressure, 2) decrease in pen velocity, and 3) rapid change in pen angle. A significant stroke is discriminated by the maximum correlation with respect to the reference signatures. Between each pair of signatures, the local correlation comparisons are computed between portions of pressure and velocity signals using segment alignment by elastic matching. Experimental results were obtained for signatures from 10 volunteers over a four-month period. The result shows that stroke based features contain robust dynamic information, and offer greater accuracy for dynamic signature verification, in comparison to results without using stroke features.

Dynamic Signature Verification using Local and Global Features

Abstract: Handwritten Signature verification is a biometric technique that is useful because signatures are in many practices accepted as a means of identity verification. This work approaches signature verification using seprate filters with different approaches. In the first, global features of the signature, such as average velocity are considered using a Euclidian distance. In the second filter, local features are considered. Strokes are segmented using the minima of the velocity and encoded before comparing them using dynamic time warping and signer-specific thresholds.

Selection of points for on-line signature comparison

Abstract: Authentication based on handwritten signature is the most accepted authentication system based on biometry because it is easy to use and because the use of signature is part of our habits. In the field of authentication by on-line signature, we present a method to reduce the amount of data to be stored for pattern comparison and that needs few processing. Many systems described in literature keep the whole signature's points even if it is not recommended, even advised against it, in order to avoid forgers to obtain a signature's pattern. The proposed method for data reduction was evaluated with respect to a method of curve comparison very often used for authentication by on-line handwritten signature: dynamic time warping (DTW). After we have presented several reduction methods of signature data, we show the results obtained with each one. In order to evaluate their efficiency, the results were compared to those obtained with the whole coarse data points of the signature.

A fast HMM algorithm based on stroke lengths for on-line recognition of handwritten music scores

Abstract: The hidden Markov model (HMM) has been successfully applied to various kinds of on-line recognition problems including, speech recognition, handwritten character recognition, etc. In this paper, we propose an on-line method to recognize handwritten music scores. To speed up the recognition process and improve usability of the system, the following methods are explained: (1) The target HMMs are restricted based on the length of a handwritten stroke, and (2) Probability calculations of HMMs are successively made as a stroke is being written. As a result, recognition rates of 85.78% and average recognition times of 5.19 ms/stroke were obtained for 6,999 test strokes of handwritten music symbols, respectively. The proposed HMM recognition rate is 2.4% higher than that achieved with the traditional method, and the processing time was 73% of that required by the traditional method.

Handling spatial information in on-line handwriting recognition

Abstract: This paper focuses on handling the two-dimensional feature of on-line handwriting signals in recognition engines. This spatial information is taken into account in various ways depending on the nature of characters to be recognized. We review some techniques used in the literature and investigate new ones to represent and model the spatial information in handwriting recognition engines. We compare formally and experimentally a number of solutions on various character recognition tasks.

Rejection strategies for offline handwritten sentence recognition

Abstract: This work investigates three different rejection strategies for offline handwritten sentence recognition. The rejection strategies are implemented as a postprocessing step of a hidden Markov model based text recognition system and are based on confidence measures derived from a list of candidate sentences produced by the recognizer. The better performing confidence measures make use of the fact that the recognizer integrates a word bigram language model. Experimental results on extracted sentences from the IAM database validate the effectiveness of the proposed rejection strategies.

Simplified Stroke-based Approach for Off-line Signature Recognition

Abstract: This paper proposes a simple adaptive off-line signature recognition method based on the feature analysis of extracted significant strokes for a given signature. Our system correctly decides on the majority of tested patterns, which include both simple and skilled forgeries. The presence of possible doubtful signatures (those ones on which is difficult to decide) is also considered. Experimental results have shown a good trade-off between response time and reasonable accuracy of recognition results.

Data hiding for multimodal biometric recognition

Abstract: Data hiding has been successfully applied lately in a variety of applications, other than copyright protection. A data hiding technique applied in a multimodal biometric system for automatic recognition is introduced in this paper. Specifically, both the voice and iris patterns acquired by an individual are hidden in certain wavelet coefficients of his fingerprint image using block classification and the embedded biometric template is further compressed. The proposed technique achieves a reduction of the system data rate, as opposed to the conventional independent compression and storage of each biometric signal. At the same time, accurate data reconstruction is guaranteed, enabling a successful recognition process based on the combination of three biometric modalities.

System and method for assessing suspicious behaviors

Abstract: Identity and biometric data are collected from individuals. The collected data is used to form biometric data packages, each of which contains information about one biometric feature and identity data associated with the corresponding individual. Each biometric data package is stored in a categorical fashion based on its biometric feature. A current query from a requesting source includes at least one monitored biometric feature of an individual of interest. A first correlation is performed between the current query's monitored biometric feature and the same type of biometric feature associated with those of the biometric data packages previously stored in a categorical fashion. A second correlation is performed between the current query's monitored biometric feature and the monitored biometric feature associated with each of the previous queries. Results of the correlations can be indications of suspicious behavior that was used by the requesting source to form the current query.

Generic target recognition

Abstract: We present in this paper a study on target recognition. The goal of this work is to determine and compare different methods from the pattern recognition domain in order to be able to recognize some objects in an image. We suppose having detected by a segmentation process a candidate object appearing with an unknown scale or rotation. To be able to recognize this object, we have first to describe it by some features having the property to be invariant by rotation, translation or scale. Second, we have to realize a supervised classification in order to compare this unknown object with one from the knowledge database. We present some experimental results for target recognition by comparing several features, classification methods and methodologies.

Unconstrained handwriting recognition: language models, perplexity, and system performance

Abstract: In this paper we present a number of language models and their behavior in the recognition of unconstrained handwritten English sentences. We use the perplexity to compare the different models and their prediction power, and relate it to the performance of a recognition system under different language models. In the recognition experiments a system with the classical architecture of preprocessing, feature extraction and recognition by means of Hidden Markov Model is used. In the recognition phase the language model constrains the possible next words. Keywords: handwriting recognition, unconstrained English sentence recognition, unigram probability, bigram probability, perplexity.

Improving DTW for Online Handwritten Signature Verification

Abstract: Authentication by handwritten on-line signature is one of the most accepted authentication system In a way, it is based on biometrics It is embedded in our cultural habits and easy to use The aim of this paper is to study two possible improvements in Dynamic Time Warping: matching process and distance computation After applying a polygonal approximation on the signatures, we test different approaches on the authentication problem Usually, the information used to match the on-line signatures are the coordinates or the speed at the input data points First, as far as the matching is concerned, we investigate other possibilities relying on the local information at each point Next, we also test several methods to take into account local information to compute the distance To limit genuine signature rejections, we use the result of the matching as a way to detect forgery and we also modify the computation of the distance Finally, we evaluate the different approaches on a base of 800 signatures The results obtained show an amelioration of the classical use of DTW

Combination of three classifiers with different architectures for handwritten word recognition

Abstract: The study of multiple classifier systems has become an area of intensive research in pattern recognition recently. Also in handwriting recognition, systems combining several classifiers have been investigated. In this paper the combination of three classifiers for handwritten word recognition with different architectures is studied. In addition a new ensemble method working with several base classifiers is applied and the results of the ensemble method are compared to the results of the combination of the three classifiers. In the experiments a large-scale handwritten word recognition task is considered.

The reduction of memory and the improvement of recognition rate for HMM on-line handwriting recognition

Abstract: The purpose of this project is two fold. The first purpose is to reduce the memory size of our previous handwriting recognition algorithm based on an HMM using self-organizing map (SOM) density tying. The second is to improve recognition capability by incorporating additional information. SOM density tying reduced the dictionary size to 1/7 of the original size, with a recognition rate of 90.45%, only slightly less than the original recognition rate of 91.51%. Our additional feature increased recognition capability to 91.34%.

Method and system for signature recognition biometrics on a fob

Abstract: The present invention discloses a system and methods for biometric security using signature recognition biometrics in a transponder-reader system. The biometric security system also includes a signature scan sensor that detects biometric samples and a device for verifying biometric samples. In one embodiment, the biometric security system includes a transponder configured with a signature scan sensor. In another embodiment, the system includes a reader configured with a signature scan sensor. In yet another embodiment, the present invention discloses methods for proffering and processing signature samples to facilitate authorization of transactions.