Resumo: Com o seqüenciamento completo do genoma do fitopatógeno Xanthomonas axonopodis pv. citri (Xac), em 2002, inúmeras possibilidades de estudo foram viabilizadas, dando margem à busca de novas formas de controle do cancro cítrico, baseadas em alvos moleculares. Estudos dessa natureza têm mostrado a existência de genes que somente são expressos quando a bactéria está se desenvolvendo in planta. Sabe-se que essa regulação é dependente da região promotora e sua identificação pode possibilitar avanços significativos na busca do controle dessa doença. Apesar do crescente avanço das técnicas experimentais in vitro em biologia molecular, identificar um número significante de promotores ainda é uma tarefa difícil e dispendiosa. Os métodos computacionais existentes enfrentam a falta de um número adequado de promotores conhecidos para identificar padrões conservados entre as espécies. Logo, um método para predizê-Ios em qualquer organismo procariótico ainda é um desafio. O Modelo Oculto de Markov é um modelo estatístico aplicável a muitas tarefas em biologia molecular. Entre elas, predição e mapeamento de seqüências promotoras no genoma de um procarioto. Neste trabalho, estudou-se o mapeamento in silico de promotores gênicos de todo o genoma da Xac e em 68% dos genes a localização de um promotor foi indicada. A análise comparativa com dados experimentais de proteômica mostrou que 72% das proteínas expressas identificaram-se com elementos desta lista de promotores, o que corresponde a 27% do total de genes do genoma. À partir destes dados foi possível levantar um rol de informações sobre estes candidatos a promotores incitando a novos estudos moleculares. / Abstract: With the complete genome sequencing of the phytopathogen Xanthomonas axonopodis pv. Citri (Xac), in 2002, several study possibilities were made practical and then creating the search of new citrus canker control ways, based in molecular aims. This kind of studies has shown the genes existences that are only expressed when the bacteria are developing itself in plant. It has been known that this regulation is promoter region dependent and its identification can allow significant advances in the search of this disease control. Although increasing advance of in vitro experimental techniques in molecular biology, identifying a meaningful number of promoters is still a hard and expensive task. The existents computer science methods face the need of a proper number of known promoters to identify conserved patterns among the species. Therefore, a method to predict them in any prokaryote organism is still a challenge. The Hidden Markov Model (HMM) is a statistic model applicable in many tasks in molecular biology. Among them, prediction and mapping of the promoters sequences in prokaryotic genome. In this work, which has studied the genic promoters in silico mapping of the whole Xac genome, in 68% of the genes the promoter localization was indicated. The proteomic experimental data comparative analysis showed that 72% of the expressed proteins identified with elements from the promoters list, which corresponds 27% of the genome genes total. According to these data it was possible to generate an information roll about these promoters candidates inciting new molecular studies. / Orientadora: Maria Inês Tiraboschi Ferro / Coorientador: Marcelo Luiz de Laia / Banca: Manoel Victor Franco Lemos / Banca: Poliana Fernanda Giachetto / Mestre

O movimento dos lábios é um recurso visual relevante para a detecção da atividade de voz do locutor e para o reconhecimento da fala. Quando os lábios estão se movendo eles transmitem a idéia de ocorrências de diálogos (conversas ou períodos de fala) para o observador, enquanto que os períodos de silêncio podem ser representados pela ausência de movimentações dos lábios (boca fechada). Baseado nesta idéia, este trabalho foca esforços para detectar a movimentação de lábios e usá-la para realizar a detecção de atividade de voz. Primeiramente, é realizada a detecção de pele e a detecção de face para reduzir a área de extração dos lábios, sendo que as regiões mais prováveis de serem lábios são computadas usando a abordagem Bayesiana dentro da área delimitada. Então, a pré-segmentação dos lábios é obtida pela limiarização da região das probabilidades calculadas. A seguir, é localizada a região da boca pelo resultado obtido na pré-segmentação dos lábios, ou seja, alguns pixels que não são de lábios e foram detectados são eliminados, e em seguida são aplicados algumas operações morfológicas para incluir alguns pixels labiais e não labiais em torno da boca. Então, uma nova segmentação de lábios é realizada sobre a região da boca depois de aplicada uma transformação de cores para realçar a região a ser segmentada. Após a segmentação, é aplicado o fechamento das lacunas internas dos lábios segmentados. Finalmente, o movimento temporal dos lábios é explorado usando o modelo das cadeias ocultas de Markov (HMMs) para detectar as prováveis ocorrências de atividades de fala dentro de uma janela temporal. / Lips motion are relevant visual feature for detecting the voice active of speaker and speech recognition. When the lips are moving, they carries an idea of occurrence of dialogues (talk) or periods of speeches to the watcher, whereas the periods of silences may be represented by the absence of lips motion (mouth closed). Based on this idea, this work focus efforts to obtain the lips motion as features and to perform visual voice activity detection. First, the algorithm performs skin segmentation and face detection to reduce the search area for lip extraction, and the most likely lip regions are computed using a Bayesian approach within the delimited area. Then, the pre-segmentation of the lips is obtained by thresholding the calculated probability region. After, it is localized the mouth region by resulted obtained in pre-segmentation of the lips, i.e., some nonlips pixels detected are eliminated, and it are applied a simple morphological operators to include some lips pixels and non-lips around the mouth. Thus, a new segmentation of lips is performed over mouth region after transformation of color to enhance the region to be segmented. And, is applied the closing of gaps internal of lips segmented. Finally, the temporal motion of the lips is explored using Hidden Markov Models (HMMs) to detect the likely occurrence of active speech within a temporal window.

The safety of miners is of interest to all countries. In the event of a coal mine disaster, how to locate the miners remains the biggest and most urgent issue. The aim of this study is to propose a precise positioning method for underground mine environments to a low cost and with acceptable accuracy. During the research work, in-depth learning and analysis of current geolocation methods for indoor areas have been carried out: advantages, disadvantages and the level of suitability of each method for mine environment have been presented. A layered two-step Hidden Markov Model has been proposed to simulate human walking in underground mine environments and an improved Viterbi algorithm suitable for the model has been implemented. The result of the positioning accuracy is quite satisfying compared to other positioning methods in the same category. A small modification to the proposed model has been illustrated in the future work which makes it more suitable for different situations rather than that limited by assumptions. The proposed positioning method can be claimed to be quite suitable for underground mine environments to a low cost and with acceptable accuracy.

Vehicular networks are an emerging technology that offer potential for providing a variety of new services. However, extending vehicular networks to include IP connections is still problematic, due in part to the incompatibility of mobile IP handovers with the increased mobility of vehicles. The handover process, consisting of discovery, registration, and packet forwarding, has a large overhead and disrupts connectivity. With increased handover frequency and smaller access point dwell times in vehicular networks, the handover causes a large degradation in performance. This thesis proposes a predictive handover solution, using a combination of a Kalman filter and an online hidden Markov model, to minimize the effects of prediction errors and to capitalize on advanced handover registration. Extensive simulated experiments were carried out in NS-2 to study the performance of the proposed solution within a variety of traffic and network topology scenarios. Results show a significant improvement to both prediction accuracy and network performance when compared to recent proposed approaches.

Les communications téléphoniques en voix sur IP souffrent de la perte de paquets causée par les problèmes d'acheminement dus aux nœuds du réseau. La perte d'un paquet de voix induit la perte d'un segment de signal de parole (généralement 10ms par paquet perdu). Face à la grande diversité des codeurs de parole, nous nous sommes intéressés dans le cadre de cette thèse à proposer une méthode de masquage de pertes de paquets générique, indépendante du codeur de parole utilisé. Ainsi, le masquage de pertes de paquets est appliqué au niveau du signal de parole reconstruit, après décodage, s'affranchissant ainsi du codeur de parole. Le système proposé repose sur une modélisation classique de type « modèles de Markov cachés » afin de suivre l'évolution acoustique de la parole. À notre connaissance, une seule étude a proposé l'utilisation des modèles de Markov cachés dans ce cadre [4]. Toutefois, Rødbro a utilisé l'utilisation de deux modèles, l'un pour la parole voisée, l'autre pour les parties non voisées, posant ainsi le problème de la distinction voisée/non voisée. Dans notre approche, un seul modèle de Markov caché est mis en œuvre. Aux paramètres classiques (10 coefficients de prédiction linéaire dans le domaine cepstral (LPCC) et dérivées premières) nous avons adjoint un nouvel indicateur continu de voisement [1, 2]. La recherche du meilleur chemin avec observations manquantes conduit à une version modifiée de l'algorithme de Viterbi pour l'estimation de ces observations. Les différentes contributions (indice de voisement, décodage acoutico-phonétique et restitution du signal) de cette thèse sont évaluées [3] en terme de taux de sur et sous segmentation, taux de reconnaissance et distances entre l'observation attendue et l'observation estimée. Nous donnons une indication de la qualité de la parole au travers d'une mesure perceptuelle : le PESQ. / Packet loss due to misrouted or delayed packets in voice over IP leads to huge voice quality degradation. Packet loss concealment algorithms try to enhance the perceptive quality of the speech. The huge variety of vocoders leads us to propose a generic framework working directly on the speech signal available after decoding. The proposed system relies on one single "hidden Markov model" to model time evolution of acoustic features. An original indicator of continuous voicing is added to conventional parameters (Linear Predictive Cepstral Coefficients) in order to handle voiced/unvoiced sound. Finding the best path with missing observations leads to one major contribution: a modified version of the Viterbi algorithm tailored for estimating missing observations. All contributions are assessed using both perceptual criteria and objective metrics.

Communication underwater is challenging because of the inherent characteristics of the media. First, common radio frequency (RF) signals utilized in wireless communications cannot be used under water. RF signals are attenuated in such as way that RF communication underwater is restricted to very few meters. As a result, acoustic-based communication is utilized for underwater communications; however, acoustic communication has its own limitations. For example, the speed of sound is five orders of magnitude lower than the speed of light, meaning that communications under water experience long propagation delays, even in short distances. Long propagation delays impose strong challenges in the design of Data Link Layer (DLL) protocols. The underwater communication channel is noisy, too. The bit error rate (BER) can also change depending on depth and other factors, and the errors are correlated, like in wireless communications. As in wireless communications, transducers for acoustic communication are half duplex, limiting the application of well-known detection mechanisms in Medium Access Control (MAC) layer protocols. Further, known problems like the hidden and exposed terminal problem also occur here. All these aspects together make the underwater communication channel to have the worst characteristics of all other known channels. Because of these reasons, underwater scenarios are complicated to implement, especially when they have underwater autonomous vehicles exchanging information among them. This dissertation proposes data link layer protocols in support of swarming of underwater autonomous vehicles that deal with the problems mentioned before. At the MAC sublayer, a MAC protocol called 2MAC is introduced. 2MAC improves the throughput of the network using the multichannel capabilities of OFDM at the physical layer. At the logical link control sublayer, a protocol named SW-MER is proposed. SW-MER improves the throughput and the reliability combining the well-known stop and wait protocol with the sliding window strategy, and using an exponential retransmission strategy to deal with errors. 2MAC and SW-MER are evaluated and compared with other protocols using analytical means and simulations. The results show that by using 2MAC, packet collisions are considerably reduced and the throughput improved. In addition, the use of SW-MER improves the packet delivery ratio over existing mechanisms. In general, the evaluations indicate that the proposed data link layer protocols offer a better communication alternative for underwater autonomous vehicles (UAV) than traditional protocols.

User behavior analysis in web applications is currently mainly performed by analysis of statistical measurements based on user interactions or by creation of personas to better understand users. Both of these methods give great insights in how the users utilize a web site, but do not give any additional information about what they are actually doing. This thesis attempts to use eye tracking data for analysis of user activities in web applications. Eye tracking data has been recorded, labeled and analyzed for 25 test participants. No data source except eye tracking data has been used and two different approaches are attempted where the first relies on a gaze map representation of the data and the second relies on sequences of features. The results indicate that it is possible to distinguish user activities in web applications, but only at a high error-rate. Improvement are possible by implementing a less subjective labeling process and by including features from other data sources. / I nuläget utförs analys av användarbeteende i webbapplikationer primärt med hjälp av statistiska mått över användares beteenden på hemsidor tillsammans med personas förökad förståelse av olika typer av användare. Dessa metoder ger stor insikt i hur användare använder hemsidor men ger ingen information om vilka typer av aktiviteter användare har utfört på hemsidan. Denna rapport försöker skapa metoder för analys av användaraktiviter på hemsidor endast baserat på blickdata fångade med eye trackers. Blick data från 25 personer har samlats in under tiden de utför olika uppgifter på olika hemsidor. Två olika tekniker har utvärderats där den ena analyserar blick kartor som fångat ögonens rörelser under 10 sekunder och den andra tekniken använder sig av sekvenser av händelser för att klassificera aktiviteter. Resultaten indikerar att det går att urskilja olika typer av vanligt förekommande användaraktiviteter genom analys av blick data. Resultatet visar också att det är stor osäkerhet i prediktionerna och ytterligare arbete är nödvändigt för att finna användbara modeller.

Cette thèse, organisée en deux parties indépendantes, a pour objet la sémantique distributionnelle et la sélection de variables. Dans la première partie, nous introduisons une nouvelle méthode pour l'apprentissage de représentations de mots à partir de grandes quantités de texte brut. Cette méthode repose sur un modèle probabiliste de la phrase, utilisant modèle de Markov caché et arbre de dépendance. Nous présentons un algorithme efficace pour réaliser l'inférence et l'apprentissage dans un tel modèle, fondé sur l'algorithme EM en ligne et la propagation de message approchée. Nous évaluons les modèles obtenus sur des taches intrinsèques, telles que prédire des jugements de similarité humains ou catégoriser des mots et deux taches extrinsèques~: la reconnaissance d'entités nommées et l'étiquetage en supersens. Dans la seconde partie, nous introduisons, dans le contexte des modèles linéaires, une nouvelle pénalité pour la sélection de variables en présence de prédicteurs fortement corrélés. Cette pénalité, appelée trace Lasso, utilise la norm trace des prédicteurs sélectionnés, qui est une relaxation convexe de leur rang, comme critère de complexité. Le trace Lasso interpole les normes $\ell_1$ et $\ell_2$. En particulier, lorsque tous les prédicteurs sont orthogonaux, il est égal à la norme $\ell_1$, tandis que lorsque tous les prédicteurs sont égaux, il est égal à la norme $\ell_2$. Nous proposons deux algorithmes pour calculer la solution du problème de régression aux moindres carrés regularisé par le trace Lasso et réalisons des expériences sur des données synthétiques. / This thesis, which is organized in two independent parts, presents work on distributional semantics and on variable selection. In the first part, we introduce a new method for learning good word representations using large quantities of unlabeled sentences. The method is based on a probabilistic model of sentence, using a hidden Markov model and a syntactic dependency tree. The latent variables, which correspond to the nodes of the dependency tree, aim at capturing the meanings of the words. We develop an efficient algorithm to perform inference and learning in those models, based on online EM and approximate message passing. We then evaluate our models on intrinsic tasks such as predicting human similarity judgements or word categorization, and on two extrinsic tasks: named entity recognition and supersense tagging. In the second part, we introduce, in the context of linear models, a new penalty function to perform variable selection in the case of highly correlated predictors. This penalty, called the trace Lasso, uses the trace norm of the selected predictors, which is a convex surrogate of their rank, as the criterion of model complexity. The trace Lasso interpolates between the $\ell_1$-norm and $\ell_2$-norm. In particular, it is equal to the $\ell_1$-norm if all predictors are orthogonal and to the $\ell_2$-norm if all predictors are equal. We propose two algorithms to compute the solution of least-squares regression regularized by the trace Lasso, and perform experiments on synthetic datasets to illustrate the behavior of the trace Lasso.

Le design du mapping (ou couplage) entre mouvement et son est essentiel à la création de systèmes interactifs sonores et musicaux. Cette thèse propose une approche appelée mapping par démonstration qui permet aux utilisateurs de créer des interactions entre mouvement et son par des exemples de gestes effectués pendant l'écoute. Le mapping par démonstration est un cadre conceptuel et technique pour la création d'interactions sonores à partir de démonstrations d'associations entre mouvement et son. L'approche utilise l'apprentissage automatique interactif pour construire le mapping à partir de démonstrations de l'utilisateur. Nous nous proposons d’exploiter la nature générative des modèles probabilistes, de la reconnaissance de geste continue à la génération de paramètres sonores. Nous avons étudié plusieurs modèles probabilistes, à la fois des modèles instantanés (Modèles de Mélanges Gaussiens) et temporels (Modèles de Markov Cachés) pour la reconnaissance, la régression, et la génération de paramètres sonores. Nous avons adopté une perspective d’apprentissage automatique interactif, avec un intérêt particulier pour l’apprentissage à partir d'un nombre restreint d’exemples et l’inférence en temps réel. Les modèles représentent soit uniquement le mouvement, soit intègrent une représentation conjointe des processus gestuels et sonores, et permettent alors de générer les trajectoires de paramètres sonores continûment depuis le mouvement. Nous avons exploré un ensemble d’applications en pratique du mouvement et danse, en design d’interaction sonore, et en musique. / Designing the relationship between motion and sound is essential to the creation of interactive systems. This thesis proposes an approach to the design of the mapping between motion and sound called Mapping-by-Demonstration. Mapping-by-Demonstration is a framework for crafting sonic interactions from demonstrations of embodied associations between motion and sound. It draws upon existing literature emphasizing the importance of bodily experience in sound perception and cognition. It uses an interactive machine learning approach to build the mapping iteratively from user demonstrations. Drawing upon related work in the fields of animation, speech processing and robotics, we propose to fully exploit the generative nature of probabilistic models, from continuous gesture recognition to continuous sound parameter generation. We studied several probabilistic models under the light of continuous interaction. We examined both instantaneous (Gaussian Mixture Model) and temporal models (Hidden Markov Model) for recognition, regression and parameter generation. We adopted an Interactive Machine Learning perspective with a focus on learning sequence models from few examples, and continuously performing recognition and mapping. The models either focus on movement, or integrate a joint representation of motion and sound. In movement models, the system learns the association between the input movement and an output modality that might be gesture labels or movement characteristics. In motion-sound models, we model motion and sound jointly, and the learned mapping directly generates sound parameters from input movements. We explored a set of applications and experiments relating to real-world problems in movement practice, sonic interaction design, and music. We proposed two approaches to movement analysis based on Hidden Markov Model and Hidden Markov Regression, respectively. We showed, through a use-case in Tai Chi performance, how the models help characterizing movement sequences across trials and performers. We presented two generic systems for movement sonification. The first system allows users to craft hand gesture control strategies for the exploration of sound textures, based on Gaussian Mixture Regression. The second system exploits the temporal modeling of Hidden Markov Regression for associating vocalizations to continuous gestures. Both systems gave birth to interactive installations that we presented to a wide public, and we started investigating their interest to support gesture learning.

No description available.