Made available in DSpace on 2015-05-14T12:36:43Z (GMT). No. of bitstreams: 1 arquivototal.pdf: 2151124 bytes, checksum: ffc486a2022781c4365766e4bf1e7054 (MD5) Previous issue date: 2009-08-27 / Coordenação de Aperfeiçoamento de Pessoal de Nível Superior / Human face has a very complex and variable pattern, which makes the face detection and recognition operations a challenging problem. The scope of these operations is quite comprehensive, involving mainly security applications, such as authorization for physical and logical access, people tracking, and real time authentication. In addition to security applications, face detection and recognition can also be associated with other applications, such as human-computer interaction and virtual reality. Several studies of face detection and recognition have been proposed and developed by researchers, pursuing greater precision and efficiency. Currently there are face detectors and recognizers with accuracy exceeding 95%. Commercial systems are available as well. This work presents a study on several face detection and recognition methods. Also was discussed the possibility of developing a new face detection method using Prediction by Partial Match (PPM), Entropy and Discrete Cosine Transform (DCT). It is further proposed a new face recognition method based on DCT. Finally, is proposed an architecture for a face detection and recognition system in video. To validate the architecture, the proposed system was implemented using one of the best detectors in the literature and the recognizer produced in this work. Several experiments were performed, and both the face detector used as the recognizer developed were effective, achieving success rates compatible with most current methods / A face humana possui um padrão bastante complexo e variável, o que torna as operações de detecção e reconhecimento de face um problema desafiador. O campo de aplicação dessas operações é bastante abrangente, envolvendo principalmente aplicações de segurança, como autorização de acesso físico e lógico, rastreamento de pessoas e autenticação em tempo real. Além de aplicações de segurança, a detecção e o reconhecimento de faces também pode ser associado a outras aplicações, como interação homem-máquina e realidade virtual. Diversos trabalhos de detecção e reconhecimento de face vêm sendo propostos e desenvolvidos pela comunidade científica, buscando continuamente uma maior precisão e eficiência. Atualmente já estão disponíveis detectores e reconhecedores de face com precisão superior a 95%. Sistemas comerciais também já estão disponíveis no mercado. Este trabalho apresenta um estudo sobre os diversos métodos de detecção e reconhecimento de face existentes. Também foi analisada a possibilidade de desenvolvimento de um novo método de detecção de face utilizando Predição por Casamento Parcial (Prediction by Partial Match, PPM), Entropia e Transformada Cosseno Discreta (Discrete Cosine Transform, DCT). Propõe-se ainda, um novo método de reconhecimento de face baseado na DCT. Por fim, apresenta-se a arquitetura de um sistema de detecção e reconhecimento de face em vídeo. Para validação desta arquitetura, o sistema proposto foi implementado utilizando um dos melhores detectores encontrados na literatura e o reconhecedor produzido neste trabalho. Diversos experimentos foram realizados e tanto o detector de face utilizado, quanto o reconhecedor desenvolvido mostraram-se eficientes, atingindo taxas de acerto compatíveis com os métodos mais atuais.

Extracting speaker-specific information from speech is of great interest to both researchers and developers alike, since speaker recognition technology finds application in a wide range of areas, primary among them being forensics and biometric security systems. Several models and techniques have been employed to extract speaker information from the speech signal. Speech production is generally modeled as an excitation source followed by a filter. Physiologically, the source corresponds to the vocal fold vibrations and the filter corresponds to the spectrum-shaping vocal tract. Vocal tract-based features like the melfrequency cepstral coefficients (MFCCs) and linear prediction cepstral coefficients have been shown to contain speaker information. However, high speed videos of the larynx show that the vocal folds of different individuals vibrate differently. Voice source (VS)-based features have also been shown to perform well in speaker recognition tasks, thereby revealing that the VS does contain speaker information. Moreover, a combination of the vocal tract and VS-based features has been shown to give an improved performance, showing that the latter contains supplementary speaker information. In this study, the focus is on extracting speaker information from the VS. The existing techniques for the same are reviewed, and it is observed that the features which are obtained by fitting a time-domain model on the VS perform poorly than those obtained by simple transformations of the VS. Here, an attempt is made to propose an alternate way of characterizing the VS to extract speaker information, and to study the merits and shortcomings of the proposed speaker-specific features. The VS cannot be measured directly. Thus, to characterize the VS, we first need an estimate of the VS, and the integrated linear prediction residual (ILPR) extracted from the speech signal is used as the VS estimate in this study. The voice source linear prediction model, which was proposed in an earlier study to obtain the ILPR, is used in this work. It is hypothesized here that a speaker’s voice may be characterized by the relative proportions of the harmonics present in the VS. The pitch synchronous discrete cosine transform (DCT) is shown to capture these, and the gross shape of the ILPR in a few coefficients. The ILPR and hence its DCT coefficients are visually observed to distinguish between speakers. However, it is also observed that they do have intra-speaker variability, and thus it is hypothesized that the distribution of the DCT coefficients may capture speaker information, and this distribution is modeled by a Gaussian mixture model (GMM). The DCT coefficients of the ILPR (termed the DCTILPR) are directly used as a feature vector in speaker identification (SID) tasks. Issues related to the GMM, like the type of covariance matrix, are studied, and it is found that diagonal covariance matrices perform better than full covariance matrices. Thus, mixtures of Gaussians having diagonal covariances are used as speaker models, and by conducting SID experiments on three standard databases, it is found that the proposed DCTILPR features fare comparably with the existing VS-based features. It is also found that the gross shape of the VS contains most of the speaker information, and the very fine structure of the VS does not help in distinguishing speakers, and instead leads to more confusion between speakers. The major drawbacks of the DCTILPR are the session and handset variability, but they are also present in existing state-of-the-art speaker-specific VS-based features and the MFCCs, and hence seem to be common problems. There are techniques to compensate these variabilities, which need to be used when the systems using these features are deployed in an actual application. The DCTILPR is found to improve the SID accuracy of a system trained with MFCC features by 12%, indicating that the DCTILPR features capture speaker information which is missed by the MFCCs. It is also found that a combination of MFCC and DCTILPR features on a speaker verification task gives significant performance improvement in the case of short test utterances. Thus, on the whole, this study proposes an alternate way of extracting speaker information from the VS, and adds to the evidence for speaker information present in the VS.

Différents types de risques existent dans le domaine financier, tels que le financement du terrorisme, le blanchiment d’argent, la fraude de cartes de crédit, la fraude d’assurance, les risques de crédit, etc. Tout type de fraude peut entraîner des conséquences catastrophiques pour des entités telles que les banques ou les compagnies d’assurances. Ces risques financiers sont généralement détectés à l'aide des algorithmes de classification. Dans les problèmes de classification, la distribution asymétrique des classes, également connue sous le nom de déséquilibre de classe (class imbalance), est un défi très commun pour la détection des fraudes. Des approches spéciales d'exploration de données sont utilisées avec les algorithmes de classification traditionnels pour résoudre ce problème. Le problème de classes déséquilibrées se produit lorsque l'une des classes dans les données a beaucoup plus d'observations que l’autre classe. Ce problème est plus vulnérable lorsque l'on considère dans le contexte des données massives (Big Data). Les données qui sont utilisées pour construire les modèles contiennent une très petite partie de groupe minoritaire qu’on considère positifs par rapport à la classe majoritaire connue sous le nom de négatifs. Dans la plupart des cas, il est plus délicat et crucial de classer correctement le groupe minoritaire plutôt que l'autre groupe, comme la détection de la fraude, le diagnostic d’une maladie, etc. Dans ces exemples, la fraude et la maladie sont les groupes minoritaires et il est plus délicat de détecter un cas de fraude en raison de ses conséquences dangereuses qu'une situation normale. Ces proportions de classes dans les données rendent très difficile à l'algorithme d'apprentissage automatique d'apprendre les caractéristiques et les modèles du groupe minoritaire. Ces algorithmes seront biaisés vers le groupe majoritaire en raison de leurs nombreux exemples dans l'ensemble de données et apprendront à les classer beaucoup plus rapidement que l'autre groupe. Dans ce travail, nous avons développé deux approches : Une première approche ou classifieur unique basée sur les k plus proches voisins et utilise le cosinus comme mesure de similarité (Cost Sensitive Cosine Similarity K-Nearest Neighbors : CoSKNN) et une deuxième approche ou approche hybride qui combine plusieurs classifieurs uniques et fondu sur l'algorithme k-modes (K-modes Imbalanced Classification Hybrid Approach : K-MICHA). Dans l'algorithme CoSKNN, notre objectif était de résoudre le problème du déséquilibre en utilisant la mesure de cosinus et en introduisant un score sensible au coût pour la classification basée sur l'algorithme de KNN. Nous avons mené une expérience de validation comparative au cours de laquelle nous avons prouvé l'efficacité de CoSKNN en termes de taux de classification correcte et de détection des fraudes. D’autre part, K-MICHA a pour objectif de regrouper des points de données similaires en termes des résultats de classifieurs. Ensuite, calculez les probabilités de fraude dans les groupes obtenus afin de les utiliser pour détecter les fraudes de nouvelles observations. Cette approche peut être utilisée pour détecter tout type de fraude financière, lorsque des données étiquetées sont disponibles. La méthode K-MICHA est appliquée dans 3 cas : données concernant la fraude par carte de crédit, paiement mobile et assurance automobile. Dans les trois études de cas, nous comparons K-MICHA au stacking en utilisant le vote, le vote pondéré, la régression logistique et l’algorithme CART. Nous avons également comparé avec Adaboost et la forêt aléatoire. Nous prouvons l'efficacité de K-MICHA sur la base de ces expériences. Nous avons également appliqué K-MICHA dans un cadre Big Data en utilisant H2O et R. Nous avons pu traiter et analyser des ensembles de données plus volumineux en très peu de temps / There are different types of risks in financial domain such as, terrorist financing, money laundering, credit card fraudulence and insurance fraudulence that may result in catastrophic consequences for entities such as banks or insurance companies. These financial risks are usually detected using classification algorithms. In classification problems, the skewed distribution of classes also known as class imbalance, is a very common challenge in financial fraud detection, where special data mining approaches are used along with the traditional classification algorithms to tackle this issue. Imbalance class problem occurs when one of the classes have more instances than another class. This problem is more vulnerable when we consider big data context. The datasets that are used to build and train the models contain an extremely small portion of minority group also known as positives in comparison to the majority class known as negatives. In most of the cases, it’s more delicate and crucial to correctly classify the minority group rather than the other group, like fraud detection, disease diagnosis, etc. In these examples, the fraud and the disease are the minority groups and it’s more delicate to detect a fraud record because of its dangerous consequences, than a normal one. These class data proportions make it very difficult to the machine learning classifier to learn the characteristics and patterns of the minority group. These classifiers will be biased towards the majority group because of their many examples in the dataset and will learn to classify them much faster than the other group. After conducting a thorough study to investigate the challenges faced in the class imbalance cases, we found that we still can’t reach an acceptable sensitivity (i.e. good classification of minority group) without a significant decrease of accuracy. This leads to another challenge which is the choice of performance measures used to evaluate models. In these cases, this choice is not straightforward, the accuracy or sensitivity alone are misleading. We use other measures like precision-recall curve or F1 - score to evaluate this trade-off between accuracy and sensitivity. Our objective is to build an imbalanced classification model that considers the extreme class imbalance and the false alarms, in a big data framework. We developed two approaches: A Cost-Sensitive Cosine Similarity K-Nearest Neighbor (CoSKNN) as a single classifier, and a K-modes Imbalance Classification Hybrid Approach (K-MICHA) as an ensemble learning methodology. In CoSKNN, our aim was to tackle the imbalance problem by using cosine similarity as a distance metric and by introducing a cost sensitive score for the classification using the KNN algorithm. We conducted a comparative validation experiment where we prove the effectiveness of CoSKNN in terms of accuracy and fraud detection. On the other hand, the aim of K-MICHA is to cluster similar data points in terms of the classifiers outputs. Then, calculating the fraud probabilities in the obtained clusters in order to use them for detecting frauds of new transactions. This approach can be used to the detection of any type of financial fraud, where labelled data are available. At the end, we applied K-MICHA to a credit card, mobile payment and auto insurance fraud data sets. In all three case studies, we compare K-MICHA with stacking using voting, weighted voting, logistic regression and CART. We also compared with Adaboost and random forest. We prove the efficiency of K-MICHA based on these experiments

Les différentes étapes et processus de la fabrication additive (FA) induisent des erreurs de sources multiples et complexes qui soulèvent des problèmes majeurs au niveau de la qualité géométrique du produit fabriqué. Par conséquent, une modélisation effective des écarts géométriques est essentielle pour la FA. Le paradigme Skin Model Shapes (SMS) offre un cadre intégral pour la modélisation des écarts géométriques des produits manufacturés et constitue ainsi une solution efficace pour la modélisation des écarts géométriques en FA.Dans cette thèse, compte tenu de la spécificité de fabrication par couche en FA, un nouveau cadre de modélisation à base de SMS est proposé pour caractériser les écarts géométriques en FA en combinant une approche dans le plan et une approche hors plan. La modélisation des écarts dans le plan vise à capturer la variabilité de la forme 2D de chaque couche. Une méthode de transformation des formes est proposée et qui consiste à représenter les effets de variations sous la forme de transformations affines appliquées à la forme nominale. Un modèle paramétrique des écarts est alors établi dans un système de coordonnées polaires, quelle que soit la complexité de la forme. Ce modèle est par la suite enrichi par un apprentissage statistique permettant la collecte simultanée de données des écarts de formes multiples et l'amélioration des performances de la méthode.La modélisation des écarts hors plan est réalisée par la déformation de la couche dans la direction de fabrication. La modélisation des écarts hors plan est effectuée à l'aide d'une méthode orientée données. Sur la base des données des écarts obtenues à partir de simulations par éléments finis, deux méthodes d'analyse modale: la transformée en cosinus discrète (DCT) et l'analyse statistique des formes (SSA) sont exploitées. De plus, les effets des paramètres des pièces et des procédés sur les modes identifiés sont caractérisés par le biais d'un modèle à base de processus Gaussien.Les méthodes présentées sont finalement utilisées pour obtenir des SMSs haute-fidélité pour la fabrication additive en déformant les contours de la couche nominale avec les écarts prédits et en reconstruisant le modèle de surface non idéale complet à partir de ces contours déformés. Une toolbox est développée dans l'environnement MATLAB pour démontrer l'efficacité des méthodes proposées. / The intricate error sources within different stages of the Additive Manufacturing (AM) process have brought about major issues regarding the dimensional and geometrical accuracy of the manufactured product. Therefore, effective modeling of the geometric deviations is critical for AM. The Skin Model Shapes (SMS) paradigm offers a comprehensive framework aiming at addressing the deviation modeling problem at different stages of product lifecycle, and is thus a promising solution for deviation modeling in AM. In this thesis, considering the layer-wise characteristic of AM, a new SMS framework is proposed which characterizes the deviations in AM with in-plane and out-of-plane perspectives. The modeling of in-plane deviation aims at capturing the variability of the 2D shape of each layer. A shape transformation perspective is proposed which maps the variational effects of deviation sources into affine transformations of the nominal shape. With this assumption, a parametric deviation model is established based on the Polar Coordinate System which manages to capture deviation patterns regardless of the shape complexity. This model is further enhanced with a statistical learning capability to simultaneously learn from deviation data of multiple shapes and improve the performance on all shapes.Out-of-plane deviation is defined as the deformation of layer in the build direction. A layer-level investigation of out-of-plane deviation is conducted with a data-driven method. Based on the deviation data collected from a number of Finite Element simulations, two modal analysis methods, Discrete Cosine Transform (DCT) and Statistical Shape Analysis (SSA), are adopted to identify the most significant deviation modes in the layer-wise data. The effect of part and process parameters on the identified modes is further characterized with a Gaussian Process (GP) model. The discussed methods are finally used to obtain high-fidelity SMSs of AM products by deforming the nominal layer contours with predicted deviations and rebuilding the complete non-ideal surface model from the deformed contours. A toolbox is developed in the MATLAB environment to demonstrate the effectiveness of the proposed methods.

This master's thesis deals with the problems of anomalies detection in data from clinical trials and medical registries. The purpose of this work is to perform literary research about quality of data in clinical trials and to design a personal algorithm for detection of anomalous records based on machine learning methods in real clinical data from current or completed clinical trials or medical registries. In the practical part is described the implemented algorithm of detection, consists of several parts: import of data from information system, preprocessing and transformation of imported data records with variables of different data types into numerical vectors, using well known statistical methods for detection outliers and evaluation of the quality and accuracy of the algorithm. The result of creating the algorithm is vector of parameters containing anomalies, which has to make the work of data manager easier. This algorithm is designed for extension the palette of information system functions (CLADE-IS) on automatic monitoring the quality of data by detecting anomalous records.

This master's thesis deals with the problems of anomalies detection in data from clinical trials and medical registries. The purpose of this work is to perform literary research about quality of data in clinical trials and to design a personal algorithm for detection of anomalous records based on machine learning methods in real clinical data from current or completed clinical trials or medical registries. In the practical part is described the implemented algorithm of detection, consists of several parts: import of data from information system, preprocessing and transformation of imported data records with variables of different data types into numerical vectors, using well known statistical methods for detection outliers and evaluation of the quality and accuracy of the algorithm. The result of creating the algorithm is vector of parameters containing anomalies, which has to make the work of data manager easier. This algorithm is designed for extension the palette of information system functions (CLADE-IS) on automatic monitoring the quality of data by detecting anomalous records.

Thesis deals with multimedia signal comparison of compression options focused on video and advanced codecs. Specifically it describes the encoding and decoding of video recordings according to the MPEG standard. The theoretical part of the thesis describes characteristic properties of the video signal and justification for the need to use recording and transmission compression. There are also described methods for elimination of encoded video signal redundancy and irrelevance. Further on are discussed ways of measuring the video signal quality. A separate chapter is focused on the characteristics of currently used and promising codecs. In the practical part of the thesis were created functions in Matlab environment. These functions were implemented into graphic user interface that simulates the activity of functional blocks of the encoder and decoder. Based on user-specified input parameters it performs encoding and decoding of any given picture, composed of images in RGB format, and displays the outputs of individual functional blocks. There are implemented algorithms for the initial processing of the input sequence including sub-sampling, as well as DCT, quantization, motion compensation and their inverse operations. Separate chapters are dedicated to the realisation of codec description in the Matlab environment and to the individual processing steps output. Further on are mentioned compress algorithm comparisons and the impact of parameter change onto the final signal. The findings are summarized in conclusion.

The biometric techniques are well known and widespread nowadays. In this context biometry means automated person recognition using anatomic features. This work uses the iris as the anatomic feature. Iris recognition is taken as the most promising technique of all because of its non-invasiveness and low error rate. The inventor of iris recognition is John G. Daugman. His work underlies almost all current public works of this technology. This final thesis is concerned with biometry based on iris images. The principles of biometric methods based on iris images are described in the first part. The first practical part of this work is aimed at the proposal and realization of two methods which localize the iris inner boundary. The third part presents the proposal and realization of iris image processing in order to classifying persons. The last chapter is focus on evaluation of experimental results and there are also compared our results with several well-known methods.

Digital image watermarking has developed for the purpose of protecting intellectual property rights to multimedia data. The focus of this thesis is searching for an alternative solution of digital image watermarking methods. A detailed analysis of watermarking methods particularly in the frequency domain, and the modification of these methods are the main aim of this work. Improved performance in watermark extraction is one of the main goals. First, the common static image watermarking methods, possible attacks on the watermarked data and techniques for objective measurement of watermarked image quality are shortly introduced. Techniques which use the space domain for watermarking ar described in the next part of this work. It is about techniques which insert the watermark into the least significant bits of an image both in the RGB domain and in the YUV domain. The main part of the thesis depicts modified and newly developed static image watermarking methods in the frequency domain. These methods use various transforms and error-correction codes, by means of which the watermark robustness increases. All the methods developed are tested in MATLAB. Results together with tables and graphs are one part of work. The end of the thesis is devoted to a comparison of all the developed methods and their evaluation.

Help documents are important in an organization to use the technology applications licensed from a vendor. Customers and internal employees frequently use and interact with the help documents section to use the applications and know about the new features and developments in them. Help documents consist of various knowledge base materials, question and answer documents and help content. In day- to-day life, customers go through these documents to set up, install or use the product. Recommending similar documents to the customers can increase customer engagement in the product and can also help them proceed without any hurdles. The main aim of this study is to build a recommendation system by exploring different machine-learning techniques to recommend the most relevant and similar help document to the user. To achieve this, in this study a hybrid-based recommendation system for help documents is proposed where the documents are recommended based on similarity of the content using content-based filtering and similarity between the users using collaborative filtering. Finally, the recommendations from content-based filtering and collaborative filtering are combined and ranked to form a comprehensive list of recommendations. The proposed approach is evaluated by the internal employees of the company and by external users. Our experimental results demonstrate that the proposed approach is feasible and provides an effective way to recommend help documents.