RCAAP Repository

Recently, in order to improve the distribution of electrical energy, and to provide agreement with the power quality standards, especial emphasis on power quality has demanded performance improvement of rectifier structures as a front end power processor. Three-phase diode-bridge rectifiers are very popular in several industrial and rural applications, where an intermediate DC link provides energy for other electronic circuits. However, such standard diode-bridge rectifiers do not meet harmonic current content restrictions, as imposed by several international standards such as IEC 61000-3-4 and IEEE 519-1992. Thus, expensive and bulky passive filters or complex power factor correction and active filter structures must be installed to compensate the inherent harmonic current distortion. Therefore, in the past few years, there has been a tremendous interest in achieving low harmonic current distortion in three-phase AC to DC converters, motivating researches of several front-end multipulse rectifiers. Several structures of multipulse rectifiers have been applied in three-phase applications for mitigation of the input current harmonic content. However, these converters need magnetic circuits such as inter-phase transformers, current balancing transformers or harmonic blocking transformers. They have complex design, are heavy, bulky, expensive and only made by special order. Besides, elimination of interphase transformers is particulary desirable when there are preexisting harmonic voltages in the three-phase power source, because those harmonic voltages cause changes in the DC output leading to further complication in the design of interphase transformers. Therefore, in this thesis, controlled rectifiers operating in parallel to standard threephase diode-bridge rectifier are proposed in contrast to expensive and complex schemes, as long as the overall converter behave as a current source controlled with a suitable strategy. The fundamental idea behind a front-end programmable rectifier is the imposition of a suitable input line current waveform in order to provide low THDI and high input power factor. In conclusion, this work proposes a new concept of multipulse hybrid rectifier for ultra clean power applications. This innovative programmable three-phase power factor correction hybrid multipulse power rectifier is capable to provide ultra clean power without the need of phase-shift transformers, inter-phase transformers, current balancing transformers or harmonic blocking transformers.

Universidade Federal de Uberlândia

Fundação de Amparo a Pesquisa do Estado de Minas Gerais

Changes in the process of teaching and learning can be both useful and nonuseful to enhance students learning. This work tried to show that current teaching methods not always meet the needs efficiently when it comes to improve students skills. For that, it relied on experiments with Language Programming subject matter s students. It did so because such a subject matter poses questions whose answers can be answered freely because its codes can be written in many ways , while its assessment is dichotomic. Teacher s didactic procedures, methodology applied in classes, programming language, and the time taken to develop the work proposed were the categories considered in the research. In an experimental group of students, Robert Mager s theory was applied to compare their learning. In such theory, instructional aims are supposed to provide a statement on the information students will get and on their understanding and ability to use them after the course ends. Conclusion is that students skills and performance were improved, that is to say, that there was an increase of knowledge, which can be measured and presented graphically by Rasch model.

Aging is directly linked to adverse impacts related to factors that affect the chronological age-related changes such as heredity, environment, diet, lifestyle, and habit of practicing physical exercises, among other features. The Central Nervous System (CNS) and neuronal signals carry information that represents changes throughout life. In this context, this study seeks to establish some correlation between brain activity as a function of age, from the record of electroencephalographic signals (EEG), in subjects not suffering from neurological disorders, while performing a certain task. There were 59 healthy subjects that voluntarily participated in this study, which were divided into 07 groups, with ages between 20 and 86 years and both sexes. EEG signals were collected \"simultaneously\" in three different experimental protocols during the execution of the Spiral of Archimedes (Ingoing Spiral, Outgoing Spiral and stopped up in the center). The electrodes were positioned according to the international standard 10/20, using the channels C3 and C4 of the central region. Statistical analyses were performed to identify differences and allow discrimination between the characteristics of each group according to the presented changes. The data were processed with software MATLAB. Among the results, significant differences were observed, via LDA-value, Linear Discriminant Analysis (LDA), a technique to optimize the extraction of discriminating information from a data set. The tool has satisfactorily performed the separation of discriminant features, classifying each group of individuals that have high correlation as a function of age. It can be concluded by the analysis of the characteristics used that there is the separability between groups according to age, contributing significantly to register the changes that occurred during the aging process.

This study presents a model for the organization of educational content customized for environments of individual studies. For many students the availability of content in general form can not be efficient. It proposed a multilevel structure of concepts to provide the development of different combinations to show the same content. The work shows that it is possible to customize the content in order to encourage other students with the use of proximal learning standards. These patterns are obtained from the analysis of the action of students with positive results in the individual organization of the content. The formal representation establishes the definition of the student profile, multi-level content, the distribution plan of correction of concepts and teaching career. The structure of the trajectory of student teaching is formally established by the method of finite differences. The system uses artificial intelligence techniques to organize and personalize content reactively. Customization is provided by an artificial neural network that enables the classification of the student profile and assign that profile to a standard proximal learning. To mediate and adjust the contents of a reactive system was inserted into a set of rules from experts in teaching. The experiment showed the applicability and appropriateness of the proposed model. The results indicated the suitability of the approach by automating the organization\'s custom content so adaptive and reactive. The intelligent system to establish the structure of the custom content to be presented was considered efficient, giving the student a better use of the content, with higher and lower final average study time and content presented.

This work demonstrates the theoretical and experimental feasibility of a induction generator to work as an attenuator of the harmonics in main generators of on the electric power system, i.e., the synchronous generators. The induction and synchronous machines were analyzed in physical terms and established mathematical modeling in the frequency domain in order to prove the purpose of the thesis. Computer simulations were also carried out in the time domain and laboratory tests were made to check levels of attenuation of current harmonics (THDi) in accordance with the maximum standardized voltage harmonic distortion (THDv). Furthermore, this work presents a study of technical and economical feasibility of a typical application.

Fundação de Amparo a Pesquisa do Estado de Minas Gerais

A person can be remembered or identified by various physical characteristics. During childhood, a child learns to identify people by voice, the smell, but probably the most remarkable characteristic of a person is face. Due to the fact that looking at the face of a person more information is collected almost simultaneously, such as eye color, mouth shape, hair texture, skin color, etc.. With a larger amount of information, the probability of incorrectly identifying a person is lower, ie it is more precise to recognize a person from the image of his face than from a voice sample. The person recognition from the image face has always been an important task. This recognition may have several objectives such as identifying a criminal, to allow people access to restricted areas, etc.. Thereforewith the improvement of computers various facial recognition techniques were developed, including techniques that use PCA (Principal Component Analysis). The purpose of this research is develop some methods of face recognition that uses less computational resources and are faster. Based on PCA method, two techniques were developed, the methods Segmented PCA and Selective PCA , which had the highest recognition rates and processing times smaller than Advanced PCA. Other three recognition techniques also developed in this thesis and theses techniques using Genetic Algorithms which also performed better than Advanced PCA. The techniques developed in this thesis can also perform facial recognition from facial regions (forehead, eyes, nose, mouth and chin), it is not always the whole picture of a person s face is available to perform the identification, for example, images obtained from security cameras where a burglar is hooded and the camera can record only the eye region.

Coordenação de Aperfeiçoamento de Pessoal de Nível Superior

An important feature in Distance Education is the personal and intelligent assistance, in which an important aspect is that students may have quite different profiles. Thus, a course designed for a particular student may not be suitable for other students. Because of this, a challenge in research is the development of advanced educational applications that provide some degree of intelligence and adaptivity. An indispensable factor in this type of system is the accurate identification of students learning styles, in order to provide most suited content to their individual needs. The construction of this kind of system through a probabilistic process is an important research problem, because these systems need to deal with incomplete or uncertain information about students. Thus, adaptivity provided by these systems need to consider a certain level of uncertainty. So, we strongly believe that considering stochastic student modeling in these systems is essential. Therefore, the main goal of this work is to present an innovative stochastic approach, based on reinforcement learning, for providing adaptivity through probabilistic modeling of students learning styles, and to investigate its efficiency, effectiveness and reliability through a series of experiments. Specifically, the proposed approach aims to detect and correct, automatically and dynamically, inaccuracies and inconsistencies in learning styles stored in the student model, considering that preferences obtained through psychometric questionnaires may need review, by having some degree of uncertainty. In this context, this work discusses and addresses important issues on automatic and dynamic modeling of learning styles, most of them ignored by approaches developed so far.

The contemporary education has a lot of challenges and among them is the adaptation of using new technologies with classical education paradigm. It hasn\'t been different with distance education. In this context, this work proposes to develop a hybrid tutoring system model with decisions based on the teachers knowledge and help from Self Organizing Maps (SOM) or Kohonen\'s Maps neural networks. The proposed system has a initial teaching method that is set by specialist teacher up and while system is being running this pedagogic method is refined by the neural networks, which use patterns extracted from students that has used the system. The model proposes the utilization of a basic neural network structure with automated training which is capable of train several networks and define the one which represents results that is more coherent with the pattern\'s set, dismissing the intervention of a specialist on the evaluation of the network training performance. The system has adaptive and reactive features related to the apprentice, being able to offer to the students a personalized and dynamic learning. The system was developed in a web environment aiming avail the advantages of this technology. At this work, besides the proposed model developing it also were performed a data gathering with fresh students from integrated learning technical of Federal Institution of Goiás, Luziânia, Goiás, Brazil, to evaluate system\'s applicability. This thesis presents the fundamentals theorists of the virtual education environment, as also the artificial neural networks SOM, used on proposed model. Likewise, it shows the system developing process, the automated training build, in addition with the system tutor structure. The knowledge\'s transmission is inspired in the content\'s didactic transposition, with organization didactic units in levels that aim develop distinct skills. The SOM networks analysis indicate that the automated train was able to train several networks and identify a network with best topologic order. Moreover, this work presents a comparison between students performance when submitted to learn using the system with purely specialized orientation and hybrid orientation. The outcomes of this evaluation points out the viability of the proposed model, since the system has shown to be able to learn from students and adapt the teacher learning method. The apprentices that studied utilizing the system had amplified theirs grades on the learning system evaluations and the hybrid tutor was capable of take decisions which magnify the acceptation of the tutor learning indication.

The face recognition has been shown to be an important technique in automatic identification of persons. It is a biometric image coding characterized by exploring selfsimilarity present in digital images and whose computational effort is significant and has required dedication of researchers in order to increasingly enhance the efficiency of the process. This thesis was presented the state of the art face recognition systems citing different forms of applications and work done by some researchers. It was also shown the methodology used by the classifiers Genetic Algorithm (GA), Artificial Neural Network (ANN) and Principal Component Analysis (PCA). The aim of this study was to test these classifiers for face recognition by analyzing the percentage of accuracy and processing time. The reason to use the GA is that the PCA, traditionally used in such cases, is very slow and the high computational cost, making it impractical in some applications, especially when the database is large images. In tests performed in this study were used the databases of two human photography files, the Olivetti Research Laboratory database today faces the British University of Cambridge (ORL) and Face Recognition Data, University of Essex, UK (UK). The accuracy rate obtained with GA, ANN and PCA were higher than those obtained by the studies reviewed.

Coordenação de Aperfeiçoamento de Pessoal de Nível Superior

The Worldwide Interoperability for Microwave Access (WiMAX) technology, based on the IEEE 802.16 standard, is a solution for broadband wireless access metropolitan networks, developed to support a wide variability of real-time and non-real time applications. Different from the traditional wireless networks, the IEEE 802.16 standard defines, in the medium access layer, mechanisms to support the Quality of Service (QoS) for the applications. Among these mechanisms, we highlight the scheduling and the Connection Admission Control (CAC). However, the IEEE 802.16 does not define the policies that must be used in the implementation of the scheduling and CAC mechanisms. The scheduling mechanism aims at guarantying the efficient utilization of the bandwidth resources, and thus, promotes the effective use of the wireless link. The CAC mechanism aims at restricting the number of existing connections simultaneously in order to avoid that the wireless link is saturated. This thesis shows a new and efficient scheduling algorithm to uplink traffic in the Base Station (BS). The proposed algorithm is developed to be totally dynamic, mainly in networks that use adaptive modulation functions. Using a cross-layer approach, a deadline based scheme was developed, aiming at minimizing the end-to-end delay for the real-time applications. Moreover, the proposed algorithm interacts with the polling mechanism of the BS, and controls the periodicity of unicast polling to real-time and non-real-time applications, in accordance with the QoS requirements of the applications. Moreover, to avoid the wireless link being saturated for an excessive number of connections, a CAC mechanism that interacts with the proposed scheduling algorithm was developed. The CAC mechanism was also developed using a cross-layer approach. Simulations results show the efficiency of the proposed scheduling algorithm and of the CAC mechanism, mainly in environments where an adaptive modulation was used.

The improvement of computer systems has benefited the development of many research areas in medicine. Analysis and interpretation of medical images represent an important part in computer vision and pattern recognition. It\'s essential for many researchers and medical centers to develop a diagnosis system aided by computer for diseases such as breast cancer, in order to assist doctors in hospitals. The polynomial classifier developed in this work is a supervised classification method and can sort two or more classes. This classifier expands the feature vector projected into the space Rd into a higher dimensional space where the classification is possible. The polynomial classifier proves to be an important method of classification mainly on treatment with non-linearly separable classes. In this thesis, the classifier was used to recognize patterns from the database IRIS Fisher 1936, which is composed by the flowers setosa, versicolor and virginica. Several tests were performed on this basis and by using three or four features the polynomial classifier managed to sort all these flowers. An application was also performed with the polynomial classifier for the identification of partial pixels in regions of interest in mammographic images. Compared with SVM classifiers and decision tree, the results were better using the polynomial classifier.

Universidade de Uberaba

The Lumbago or commonly called "low back pain" is a frequent cause of morbidity and disability work, only surpassed by headache (headache) in the range of painful disorders affecting modern society, and the treatment is based on medication (painkillers) to relieve the symptoms of pain and recommending therapeutic exercise. However, when these exercises are performed and no longer perceived pain, the patient drops the treatment, and in many cases, having its range of movement fully reestablished. To mitigate this index abandoned therapies motivators several techniques have been employed, such as the use of digital games and entertainment interfaces. However, these approaches only emerge the playful part of treatment showing no concrete data on the occupational or even exercise performance and its efficiency in treatment. Therefore, this work describes a rehabilitation system based on motion capture techniques for the treatment of a type of low back pain, mechanical low back pain. This thesis presents techniques use motion capture environments applied to motor rehabilitation with some devices such as electrogoniometer, extensometer, accelerometer and other common interaction devices such as webcams, video game console (Nintendo WiiTM) and motion sensor MS-KinectTM. The motivation of the work is based on the fact that motion capture systems for video devices tend to be more economically viable, however the systems capture current video showed the following limitations: the overlap and occlusion of the markers, the disparity optical cameras the field of view and their geometric constraints. Therefore, this work presents an alternative proposal capture by multiple cameras in conjunction with the analysis of the intrinsic parameters of the cameras involved in the process to mitigate these limitations. For the analysis of motion captured was prototyped a system that meets the specifications on the biotype of the patient and an adapted questionnaire about their inability to move that, from these, form a set of information that supports the decision about treatment with exercise suggestions . In order to validate the information in the system as well as your business rules, an evaluation has made with a group of five therapists, and presented an analysis of the interface has performed by 20 patients. Those patients were surveyed volunteers. Through this analysis was possible to identify the potential of the prototype, as well as highlight the strategies for implementing even small physiotherapy clinics. Especially those that have no spaces or volume of investment in infrastructure for systems acquisition motion capture large - very important for analysis and recommendation of new exercises during a rehabilitative therapy.

Conselho Nacional de Desenvolvimento Científico e Tecnológico

Learning music, mostly through musical scores, is a difficult task, even for people who do not have visual problems. Learning musical demands years of effort and dedication. Moreover, the music learning tasks for a blind is much more complex. Besides the complexity of the Braille code for music (musicografia Braille), there are few people trained to teach music by this method. This work presents a solution that seeks to simplify the coding of both writing and reading music through cells of raised dots, as also make teaching and learning more attractive, requiring visually impaired and their assistants, less effort, less time and less dedication. This work presents a computational solution that seeks to give the visually impaired, accessibility to the songs as they are posted on the same internet. Unfortunately, the biggest barrier to be faced, it is still a lack of professionals to teach the blind to read such encoding. The proposed system, prototyped and now presents a new simplified coding, presenting the music score in basic cells of 4 points in relief (4 raised dots arranged in cells), musical, textual and Numerofonia format. The proposed method is based on the paradigm of decimal encoding proposed by Braille and the paradigm number already used in international encoding of musical events. The method of this research presents a paradigm that allows, with less effort and less dedication, training of teachers, family members and assistants of visually impaired in learning musical notation proposed here. Learning to read music is relevant due to the fact that, since the assistant learn the paradigm and he can read music, it will be easier to teach visually impaired under his tutelage, the new musicografia. Software was designed and implemented in functional language for generating the codes of this work, as well as other paradigms focused on use by accompanying of visually impaired.