Coordenação de Aperfeiçoamento de Pessoal de Nível Superior - CAPES / This work is a study of the relationship between the intrinsic dimension of feature vectors applied to the classification of video signals in order to perform lip reading. In pattern recognition tasks, the extraction of relevant features is crucial for a good performance of the classifiers. The starting point of this work was the reproduction of the work of J.R. Movellan [1], which classifies lips gestures with HMM using only the video signal from the Tulips1 database. The database consists of videos of volunteers’ mouths while they utter the first 4 numerals in English. The original work uses feature vectors of high dimensionality in relation to the size of the database. Consequently, the adjustment of HMM classifiers has become problematic and the maximum accuracy was only 66.67%. Alternative strategies for feature extraction and classification schemes were proposed in order to analyze the influence of the intrinsic dimension in the performance of classifiers. The best solution, in terms of results, achieved an accuracy of approximately 83%. / Este trabalho é um estudo da relação entre a dimensão intrínseca de vetores de características aplicados à classificação de sinais de vídeo no intuito de realizar-se a leitura labial. Nas tarefas de reconhecimento de padrões, a extração de características relevantes é crucial para um bom desempenho dos classificadores. O ponto de partida deste trabalho foi a reprodução do trabalho de J.R. Movellan [1], que realiza a classificação de gestos labiais com HMM na base de dados Tulips1, utilizando somente o sinal de vídeo. A base é composta por vídeos das bocas de voluntários enquanto esses pronunciam os primeiros 4 numerais em inglês. O trabalho original utiliza vetores de características de dimensão muito alta em relação ao tamanho da base. Consequentemente, o ajuste de classificadores HMM se tornou problemático e só se alcançou 66,67% de acurácia. Estratégias de extração de características e esquemas de classificação alternativos foram propostos, a fim de analisar a influência da dimensão intrínseca no desempenho de classificadores. A melhor solução, em termos de resultados, obteve uma acurácia de aproximadamente 83%. / São Cristóvão, SE

Submitted by Claudinei Pracidelli (cpracide@ipen.br) on 2016-11-11T09:45:23Z No. of bitstreams: 0 / Made available in DSpace on 2016-11-11T09:45:23Z (GMT). No. of bitstreams: 0 / Um desafio recorrente em processos produtivos é o desenvolvimento de sistemas de monitoração e diagnóstico. Esses sistemas ajudam na detecção de mudanças inesperadas e interrupções, prevenindo perdas e mitigando riscos. Redes Neurais Artificiais (RNA) têm sido largamente utilizadas na criação de sistemas de monitoração. Normalmente as RNA utilizadas para resolver este tipo de problema são criadas levando-se em conta apenas parâmetros como o número de entradas, saídas e quantidade de neurônios nas camadas escondidas. Assim, as redes resultantes geralmente possuem uma configuração onde há uma total conexão entre os neurônios de uma camada e os da camada seguinte, sem que haja melhorias em sua topologia. Este trabalho utiliza o algoritmo de Otimização por Colônia de Formigas (OCF) para criar redes neurais otimizadas. O algoritmo de busca OCF utiliza a técnica de retropropagação de erros para otimizar a topologia da rede neural sugerindo as melhores conexões entre os neurônios. A RNA resultante foi aplicada para monitorar variáveis do reator de pesquisas IEA-R1 do IPEN. Os resultados obtidos mostram que o algoritmo desenvolvido é capaz de melhorar o desempenho do modelo que estima o valor de variáveis do reator. Em testes com diferentes números de neurônios na camada escondida, utilizando como comparativos o erro quadrático médio, o erro absoluto médio e o coeficiente de correlação, o desempenho da RNA otimizada foi igual ou superior ao da tradicional. / Dissertação (Mestrado em Tecnologia Nuclear) / IPEN/D / Instituto de Pesquisas Energeticas e Nucleares - IPEN-CNEN/SP

Submitted by Alisson Mota (alisson.davidbeckam@gmail.com) on 2015-07-14T20:42:42Z No. of bitstreams: 1 Dissertação - Marinete Lourenço Mota.pdf: 2482577 bytes, checksum: a54f67d62bbe1d1231580464a8935bbc (MD5) / Approved for entry into archive by Divisão de Documentação/BC Biblioteca Central (ddbc@ufam.edu.br) on 2015-07-15T20:09:30Z (GMT) No. of bitstreams: 1 Dissertação - Marinete Lourenço Mota.pdf: 2482577 bytes, checksum: a54f67d62bbe1d1231580464a8935bbc (MD5) / Approved for entry into archive by Divisão de Documentação/BC Biblioteca Central (ddbc@ufam.edu.br) on 2015-07-15T20:18:02Z (GMT) No. of bitstreams: 1 Dissertação - Marinete Lourenço Mota.pdf: 2482577 bytes, checksum: a54f67d62bbe1d1231580464a8935bbc (MD5) / Made available in DSpace on 2015-07-15T20:18:02Z (GMT). No. of bitstreams: 1 Dissertação - Marinete Lourenço Mota.pdf: 2482577 bytes, checksum: a54f67d62bbe1d1231580464a8935bbc (MD5) Previous issue date: 2010-11-12 / CAPES - Coordenação de Aperfeiçoamento de Pessoal de Nível Superior / Hunt is seen as an activity of killing or capture of animal life elements, practiced since the beginning of times where man already used animals in different ways as food, clothing, medicine or pet animals. In the Amazon forest this activity is practiced throughout its social formation being considered source of protein to sustenance of populations that survive in it. With the objective of investigate the strategies of hunt and relations between fauna and people that live in the community of Pini, in the National Forest of Tapajós, we characterized the areas of hunt recognized as stains by the local hunters, as were used photographic traps and semi-structured interviews to register the forest fauna to know what animals are consumed and in witch quantities by the local community members with the intent of cooperate with the consumption plan of the unit that is being studied by the federal university of Pará – UFPA, through the Amazon high studies nucleus – NAEA, institution that is searching conservation measures that guarantees the prudent and sustainable consumption of the available resources to that community. So there were identified three types of forests: firm, Capoeira and Igapó, and in all these locations there were (comidias) fruit trees. Through the photographic traps there were registered 10 species of animals, all mammals: armadillo (Dasypus novemcinctus), cutia (Dasyprocta agouti), deer (Mazama americana), irara (Eira barbara), mucura (Didelphis marsupialis) and to birds: mutum (Mitu mitu), red juriti (Geotrygon violaceae), gavião (Morphnus guianensis), inambu (Tinamus tao) e jacamim (Psophia creptans). Between hunts, the most registered were the one the hunter waits over a platform and the shotgun the most used. Fish was the most consumed food by the local residents, followed by hunt and among the registered animals between the semi-structured interviews, direct and indirect observations were registered 16 species of animals distributed in six orders and 25 families. Among all those animals, the armadillo, cutia and inambu were registered in all the techniques of sample and the most abundant mammals were paca, armadillo and cutia and to birds: mutum and inambu, and the most abundant reptile was the tortoise (Podocnemis expansa). The study revealed that even not having evidences about over exploitation of hunt, we can see that the hunted fauna is being found in more distant places, therefore, the exploration of hunt in this community can be as an economic alternative of production of low impact by the social and economic point of view, being able to exist of this animal resources among the local social participation and these villagers can demand from ICMBio measures that guarantee the autonomy of the community, since the participative plan and society involvement in planning suggests measures of evaluation and monitoring of hunt exploitation by the competent authorities and this will be of fundamental importance to the maintenance of the hunt activity to sustainable of this community off the national forest of Tapajós. / A caça é entendida como atividade de abate ou captura de elementos da vida animal, praticada desde os primórdios onde homem já utilizava esta fauna de diversas maneiras como alimentação, vestimenta, medicamentos ou animal de estimação. Na floresta Amazônica essa atividade é prática envolvida na história de sua formação social sendo considerada fonte de proteína para subsistência de populações que nela sobrevivem. Com o objetivo de investigar as estratégias de caça e as relações entre a fauna e pessoas que vivem na comunidade de Pini na Floresta Nacional do Tapajós, foi feita a caracterização das áreas de caça reconhecidas como manchas pelos caçadores locais, assim como foram utilizadas armadilhas fotográficas e entrevistas semi-estruturadas para registrar a fauna silvestre para saber quais animais são consumidos e em quais quantidades pelos moradores da comunidade com o intuito de colaborar com o plano de manejo da área que está em estudo pela Universidade Federal do Pará – UFPA, através do Núcleo de Altos Estudos amazônicos – NAEA, instituição que está buscando medidas de conservação que garanta o manejo prudente e sustentável dos recursos disponíveis para esta comunidade. Assim, foram identificados três tipos florestais:Terra Firme, Capoeira e Igapó e em todos esses locais havia (comidias) fruteiras. Através das armadilhas fotográficas foram registradas 10 espécies de animais sendo os mamíferos: tatu (Dasypus novemcinctus), cutia (Dasyprocta agouti), veado (Mazama americana), irara (Eira barbara), mucura (Didelphis marsupialis) e para as aves: mutum (Mitu mitu), juriti vermelha (Geotrygon violaceae), gavião (Morphnus guianensis), inambu (Tinamus tao) e jacamim (Psophia creptans). Entre as caçadas, as mais registradas foram de espera com a espingarda sendo a arma mais usada. O peixe foi o alimento mais consumido pelos moradores seguido da caça e dentre os animais registrados entre as entrevistas semi-estruturadas, observações diretas e indiretas foram registradas 16 espécies de animais distribuídos em seis ordens e 25 famílias. Entre todos esses animais, as espécies tatu, cutia e inambu foram registrados em todas as técnicas de amostragens e os mamíferos mais abundantes foram: paca, tatu e cutia e para as aves: mutum e inambu e o réptil mais abundante foi tartaruga (Podocnemis expansa). O estudo revelou que mesmo não tendo evidências de sobre-exploração da caça, percebe-se que a fauna caçada está sendo encontrados em manchas mais distantes, portanto, o manejo de caça nesta comunidade pode servir como alternativa econômica de produção de impacto reduzido do ponto de vista social e econômico, podendo existir sustentabilidade desses recursos animais junto à participação social local e esses moradores podem exigir do ICMBio medidas que garantam a autonomia da comunidade, já que o plano é participativo e o envolvimento da sociedade no planejamento sugere medidas de avaliação e fiscalização da exploração de caça pelo órgão competente e isso será de fundamental importância para a manutenção da atividade de caça a níveis sustentáveis desta comunidade da Floresta Nacional do Tapajós.

Submitted by Cássia Santos (cassia.bcufg@gmail.com) on 2017-02-22T11:00:53Z No. of bitstreams: 2 Dissertação - Spiridon Nicofotis Anyfantis - 2004.pdf: 1754042 bytes, checksum: 3be2f4d514c73ade66e257c0e0b50294 (MD5) license_rdf: 0 bytes, checksum: d41d8cd98f00b204e9800998ecf8427e (MD5) / Approved for entry into archive by Luciana Ferreira (lucgeral@gmail.com) on 2017-02-22T13:17:29Z (GMT) No. of bitstreams: 2 Dissertação - Spiridon Nicofotis Anyfantis - 2004.pdf: 1754042 bytes, checksum: 3be2f4d514c73ade66e257c0e0b50294 (MD5) license_rdf: 0 bytes, checksum: d41d8cd98f00b204e9800998ecf8427e (MD5) / Made available in DSpace on 2017-02-22T13:17:29Z (GMT). No. of bitstreams: 2 Dissertação - Spiridon Nicofotis Anyfantis - 2004.pdf: 1754042 bytes, checksum: 3be2f4d514c73ade66e257c0e0b50294 (MD5) license_rdf: 0 bytes, checksum: d41d8cd98f00b204e9800998ecf8427e (MD5) Previous issue date: 2004-03-02 / It concerns an essay about the evidences in the penal procedure and its way of production by means of image captation and recordings, whether by hidden cameras or safety ones. It´s composed of five chapters which analyses, respectively, the penal evidence in its particularities and afterwards, the theme of the illegal evidences making, inclusively, a counterpoint between proof illicitness and the proportionateness principle. It still studies the theme of exceptions to the theory of illegal proofs – named by argentinean and spanish exception of the exclusionary rule. They´re studied, briefly, intimacy aspects and individual private life, with subtle distinctions and similarities. Soon after it goes to an analyses regarding the original image recordings evidence – by ostensive cameras and hidden ones – and its use in the penal procedure, specially when used in work places, investigative journalism, private and public places among other circumstances. Finally, such images validity hypothesis are evaluated in the penal procedure and suggestions are presented in order to them, in case of being produced, and in very specific cases, can be effectively used in the process without violation to the intimacy and private life of the individual. / Trata-se de estudo sobre a prova no processo penal e sua forma de produção por meio de captação e gravação de imagens, seja por câmeras ocultas ou de segurança. É composto por cinco capítulos onde se analisam, respectivamente, a prova penal em suas particularidades e, em seguida, o tema das provas ilícitas, fazendo-se, inclusive, contraponto entre a ilicitude da prova e o princípio da proporcionalidade. É estudado, ainda, o tema das exceções à teoria das provas ilícitas – denominada pelos argentinos e espanhóis exceções à regra de exclusão probatória. São estudados, sucintamente, aspectos da intimidade e da vida privada do indivíduo, com suas sutis distinções e semelhanças. É analisada a questão das provas originárias de gravações de imagens – por câmeras ostensivas e ocultas - e sua utilização no processo penal, especialmente quando utilizadas em ambientes de trabalho, jornalismo investigativo, ambientes privados e públicos, dentre outras circunstâncias. São avaliadas as hipóteses de validade de tais imagens no processo penal e apresentadas sugestões para que as mesmas, caso produzidas, e em hipóteses muito específicas, possam ser efetivamente utilizadas no processo sem violação à intimidade e vida privada do indivíduo.

The aim of this thesis is to compare the different Crouching Tiger, Hidden Dragon film reviews from Chinese and American reviewers respectively, and to see how they evaluate and understand the film in the different ways. We use reception theory as our theoretical tool, and divide the film into four themes so that we can compare the different reviews in a more relative way. Meanwhile, we also present some brief ideas to explain the reason why the film gets the different response from both sides. Since our main purpose is to compare the different film reviews, so we choose the comparative design as our research design. The main findings of this thesis show that Chinese and American reviewers do have some distinct evaluations, understandings and views about the film.Besides, culture and aesthetic become the two major factors to influence the response from both sides.

Single molecule experiments study the kinetics of molecular biological systems. Many such studies generate data that can be described by aggregated hidden Markov models, whereby there is a need of doing inference on such data and models. In this study, model selection in aggregated Hidden Markov models was performed with a criterion of maximum Bayesian evidence. Variational Bayes inference was seen to underestimate the evidence for aggregated model fits. Estimation of the evidence integral by brute force Monte Carlo integration theoretically always converges to the correct value, but it converges in far from tractable time. Nested sampling is a promising method for solving this problem by doing faster Monte Carlo integration, but it was here seen to have difficulties generating uncorrelated samples.

Kyoto University (京都大学) / 0048 / 新制・課程博士 / 博士(工学) / 甲第14919号 / 工博第3146号 / 新制||工||1472(附属図書館) / 27357 / UT51-2009-M833 / 京都大学大学院工学研究科都市社会工学専攻 / (主査)教授 小林 潔司, 教授 大津 宏康, 教授 河野 広隆 / 学位規則第4条第1項該当

Abstract Medium access control (MAC) in wireless ad hoc networks has received considerable attention for almost a couple of decades; however, there are still open problems which deserve thorough study in order to facilitate migration to the next generation broadband wireless communication systems. In ad hoc networks, a detected frame collision can be due to the so-called unreachability problem, where the destination station is situated either in the transmission or interference range of an emitting station and is unable to receive connection establishment frames from any of its neighboring stations. Unreachability might also be due to the inability of a radio station to respond to any connection establishment request, though when the unreachable station receives the connection establishment requests, however, it is prohibited from responding to the requests due to being situated in the interference range of the emitting neighbor. To investigate the impact of this problem, we have to be equipped with a proper analytical framework; therefore, as the first part of this thesis, a scalable framework called Parallel Space – Time Markov chain (PSTMC) is proposed, through which a finite load non-saturated ad hoc network can be easily modeled. At the first step, a single-hop ad hoc network is considered and the accuracy of the model is evaluated using extensive numerical results. Subsequently, the proposed framework is further extended to model multi-hop ad hoc networks. Several discussions are also given on how the framework can be deployed for an arbitrary network topology. One of the main key features of the PSTMC model is its remarkable scalability in modeling complex network configurations. In fact, it is shown that multi-hop ad hoc networks have bounded complexity in being modeled by the PSTMC framework due to its spectacular specifications. These features lead us to a powerful tool by which an arbitrary network topology can be studied. In addition, the proposed models clearly facilitate demonstrating the impact of the unreachability problem on the performance of multi-hop networks. The introduced framework shows how the unreachability problem degrades the achieved throughput and channel capacity by the contending radio stations depending on the deployed network topology. In the remainder of the thesis the unreachability problem in mobile ad hoc networks is tackled and a new MAC protocol to enhance the performance of the network is proposed. This MAC scheme is equipped with smart decision-making algorithms as well as adaptive management mechanisms to reduce the impact of the unreachability problem in single channel scenarios. Subsequently, the problem of concurrent radio resource management and contention resolution in multi-channel cognitive ad hoc networks is considered. In particular, a multi-channel technique for traffic distribution among a set of data channels without centralized control, which is enabled by a probabilistic channel selection algorithm as well as a multi-channel binary exponential backoff mechanism, is proposed. It is shown through simulations that the suggested scheme outperforms the existing MAC protocols in multi-channel environments as well as cognitive networks coexisting with primary users. A mathematical model is also introduced to study the performance of the multi-channel MAC protocol in a single-hop non-saturated wireless network.

The vital signs monitoring data of an infant receiving intensive care are a rich source of information about its health condition. One major concern about the state of health of such patients is the onset of neonatal sepsis, a life-threatening bloodstream infection. As early signs are subtle and current diagnosis procedures involve slow laboratory testing, sepsis detection based on the monitored physiological dynamics is a clinically significant task. This challenging problem can be thoroughly modelled as real-time inference within a machine learning framework. In this thesis, we develop probabilistic dynamical models centred around the goal of providing useful predictions about the onset of neonatal sepsis. This research is characterised by the careful incorporation of domain knowledge for the purpose of extracting the infant’s true physiology from the monitoring data. We make two main contributions. The first one is the formulation of sepsis detection as learning and inference in an Auto-Regressive Hidden Markov Model (AR-HMM). The model investigates the extent to which physiological events observed in the patient’s monitoring traces could be used for the early detection of neonatal sepsis. In addition, the proposed approach involves exact marginalisation over missing data at inference time. When applying the ARHMM on a real-world dataset, we found that it can produce effective predictions about the onset of sepsis. Second, both sepsis and clinical event detection are formulated as learning and inference in a Hierarchical Switching Linear Dynamical System (HSLDS). The HSLDS models dynamical systems where complex interactions between modes of operation can be represented as a twolevel hidden discrete hierarchical structure. For neonatal condition monitoring, the lower layer models clinical events and is controlled by upper layer variables with semantics sepsis/nonsepsis. The model parameterisation and estimation procedures are adapted to the specifics of physiological monitoring data. We demonstrate that the performance of the HSLDS for the detection of sepsis is not statistically different from the AR-HMM, despite the fact that the latter model is given “ground truth” annotations of the patient’s physiology.

Philosophiae Doctor - PhD / In this thesis, we explore dynamic signature verification systems. Unlike other signature models, we use genuine signatures in this project as they are more appropriate in real world applications. Signature verification systems are typical examples of biometric devices that use physical and behavioral characteristics to verify that a person really is who he or she claims to be. Other popular biometric examples include fingerprint scanners and hand geometry devices. Hand written signatures have been used for some time to endorse financial transactions and legal contracts although little or no verification of signatures is done. This sets it apart from the other biometrics as it is well accepted method of authentication. Until more recently, only hidden Markov models were used for model construction. Ongoing research on signature verification has revealed that more accurate results can be achieved by combining results of multiple models. We also proposed to use combinations of multiple single variate models instead of single multi variate models which are currently being adapted by many systems. Apart from these, the proposed system is an attractive way for making financial transactions more secure and authenticate electronic documents as it can be easily integrated into existing transaction procedures and electronic communications