José F. Vélez

SUPPORT VECTOR MACHINES VERSUS MULTI-LAYER PERCEPTRONS FOR EFFICIENT OFF-LINE SIGNATURE RECOGNITION

Abstract The problem of automatic signature recognition has received little attention in comparison with the problem of signature verification despite its potential applications for accessing security-sensitive facilities and for processing certain legal and historical documents. This paper presents an efficient off-line human signature recognition system based on support vector machines (SVM) and compares its performance with a traditional classification technique, multi-layer perceptrons (MLP). In both cases we propose two approaches to the problem: (1) construct each feature vector using a set of global geometric and moment-based characteristics from each signature and (2) construct the feature vector using the bitmap of the corresponding signature. We also present a mechanism to capture the intrapersonal variability of each user using just one original signature. Our results empirically show that SVM, which achieves up to 71% correct recognition rate, outperforms MLP.

Robust off-line signature verification using compression networks and positional cuttings

A novel robust technique for the off-line signature verification problem in practical real conditions is presented. The technique is based on the use of compression neural networks, and in the automatic generation of the training set from only one signature for each writer. Our proposal incorporates a new kind of acceptance/rejection rule, which is based on the similarity between subimages or positional cuttings of a test signature and the corresponding representation stored in the class compression network. Experimental results show that the proposed technique reduces significantly the false acceptation rate (FAR).

Convolutional Neural Networks and Long Short-Term Memory for skeleton-based human activity and hand gesture recognition

Combination of a Convolutional Neural Network (CNN) and a Long Short-Term Memory (LSTM) recurrent network for skeleton-based human activity and hand gesture recognition.Two-stage training strategy which firstly focuses on the CNN training and, secondly, adjusts the full method CNN+LSTM.A method for data augmentation in the context of spatiotemporal 3D data sequences.An exhaustive experimental study on publicly available data benchmarks with respect to the state-of-the-art most representative methods.Comparison among different CPU and GPU platforms. In this work, we address human activity and hand gesture recognition problems using 3D data sequences obtained from full-body and hand skeletons, respectively. To this aim, we propose a deep learning-based approach for temporal 3D pose recognition problems based on a combination of a Convolutional Neural Network (CNN) and a Long Short-Term Memory (LSTM) recurrent network. We also present a two-stage training strategy which firstly focuses on CNN training and, secondly, adjusts the full method (CNN+LSTM). Experimental testing demonstrated that our training method obtains better results than a single-stage training strategy. Additionally, we propose a data augmentation method that has also been validated experimentally. Finally, we perform an extensive experimental study on publicly available data benchmarks. The results obtained show how the proposed approach reaches state-of-the-art performance when compared to the methods identified in the literature. The best results were obtained for small datasets, where the proposed data augmentation strategy has greater impact.

Analyzing the influence of contrast in large-scale recognition of natural images

This paper analyzes both the isolated influence of illumination quality in 2D facial recognition and also the influence of contrast measures in large-scale recognition of low-resolution natural images. First, using the Yale Face Database B, we have shown that by separately estimating the illumination quality of facial images (through a fuzzy inference system that combines av- erage brightness and global contrast of the patterns) and by recognizing the same images using a multilayer perceptron, there ex- ists a nearly-linear correlation between both illumination and recognition results. Second, we introduced a new contrast measure, called Harris Points Measured Contrast (HPMC), which assigns values of contrast in a more consistent form to images, according to their recognition rate than other global and local compared contrast analysis methods. For our experiments on image contrast analysis, we have used the CIFAR-10 dataset with 60,000 images and convolutional neural networks as classification models. Our results can be considered to decide if it is worth using a given test image, according to its calculated contrast applying the proposed HPCM metric, for further recognition tasks.

Three-dimensional facial surface modeling applied to recognition

Applications related to game technology, law-enforcement, security, medicine or biometrics are becoming increasingly important, which, combined with the proliferation of three-dimensional (3D) scanning hardware, have made that 3D face recognition is now becoming a promising and feasible alternative to two-dimensional (2D) face methods. The main advantage of 3D data, when compared with traditional 2D approaches, is that it provides information that is invariant to rigid geometric transformations and to pose and illumination conditions. One key element for any 3D face recognition system is the modeling of the available scanned data. This paper presents new 3D models for facial surface representation and evaluates them using two matching approaches: one based on support vector machines and another one on principal component analysis (with a Euclidean classifier). Also, two types of environments were tested in order to check the robustness of the proposed models: a controlled environment with respect to facial conditions (i.e. expressions, face rotations, etc.) and a non-controlled one (presenting face rotations and pronounced facial expressions). The recognition rates obtained using reduced spatial resolution representations (a 77.86% for non-controlled environments and a 90.16% for controlled environments, respectively) show that the proposed models can be effectively used for practical face recognition applications.

INTRODUCING ALGORITHM DESIGN TECHNIQUES IN UNDERGRADUATE DIGITAL IMAGE PROCESSING COURSES

This paper documents the development and first offering of an undergraduate course in Digital Image Processing at the Rey Juan Carlos University, Madrid (Spain). The paper describes how the appropriate introduction of main Algorithm Design Techniques can successfully assist the students to achieve a comprehensive understanding of image operations and related algorithms. Image processing problems offer a natural way to present real world problems where the students can use their algorithmic knowledge. Furthermore, image processing solutions are needed from a methodological development and require efficient well-designed algorithms. This paper presents an effort in the integration of Algorithm Design Techniques in a Digital Image Processing course with a very practical scope.

Off-line handwritten signature detection by analysis of evidence accumulation

One fundamental step in off-line handwritten signature verification is the detection of the signature position within the document image. This paper introduces an original approach for signature position detection. The method is based on an accumulative evidence technique, searching the region that maximizes some measure of correspondence with a given reference signature. This measure is based on the similarity of the slope marked out by each of the strokes in the signature. Experiments have shown that the method can be used on real documents, such as bank checks, where images have a high noise level due to background interferences (i.e. machine or handwritten texts, stamps, and lines). The proposed method is robust to variability in the size of the signatures and has the advantage of using only one reference signature per person.

Introducing Fuzziness on Snake Models for Off-Line Signature Verification: A Comparative Study

This paper presents an experimental comparison of different hybrid snake-based algorithms for automatic off-line signature verification. Snakes are usually applied to different image analysis tasks, especially to segmentation. Off-line signature verification aims to establish the degree of genuineness between one given test signature and one reference signature. Due to the intrapersonal variability in human signatures, a system with tolerance to imprecision seems be appropriate for this verification task. Our work aims to study how effective is introducing fuzziness to tackle this image analysis problem. We developed several hybrid snake algorithms adapted to the practical requirements of automatic signature verification. Fuzziness is introduced in the feature extraction stage and during the signature verification (or classification) stage. Our work compares four different hybrid snake-based approaches for this verification problem using the corresponding FAR and FRR biometric errors. Experimental results have shown that none of the tested approaches clearly outperforms the other ones.

Gender and Handedness Prediction from Offline Handwriting Using Convolutional Neural Networks

Demographic handwriting-based classification problems, such as gender and handedness categorizations, present interesting applications in disciplines like Forensic Biometrics. This work describes an experimental study on the suitability of deep neural networks to three automatic demographic problems: gender, handedness, and combined gender-and-handedness classifications, respectively. Our research was carried out on two public handwriting databases: the IAM dataset containing English texts and the KHATT one with Arabic texts. The considered problems present a high intrinsic difficulty when extracting specific relevant features for discriminating the involved subclasses. Our solution is based on convolutional neural networks since these models had proven better capabilities to extract good features when compared to hand-crafted ones. Our work also describes the first approach to the combined gender-and-handedness prediction, which has not been addressed before by other researchers. Moreover, the proposed solutions have been designed using a unique network configuration for the three considered demographic problems, which has the advantage of simplifying the design complexity and debugging of these deep architectures when handling related handwriting problems. Finally, the comparison of achieved results to those presented in related works revealed the best average accuracy in the gender classification problem for the considered datasets.

A method for K-Means seeds generation applied to text mining

In this paper, a methodology is proposed in order to produce a set of seeds later used as a starting point to K-Means-type unsupervised classification algorithms for text mining. Our proposal involves using the eigenvectors obtained from principal component analysis to extract initial seeds, upon appropriate treatment for search of lightly overlapping clusters which are also clearly identified by keywords. This work is motivated by the interest of the authors in the problem of identification of topics and themes previously unknown in short texts. Therefore, in order to validate the goodness of this method, it was applied on a sample of labeled e-mails (NG20) representing a gold standard within the field of text mining. Specifically, some corpora referenced in the literature have been used, configured in accordance to a mix of topics contained in the sample. The proposed method improves on the results of other state-of-the-art methods to which it is compared.