Dissertations / Theses on the topic 'Visemes'

This thesis introduces a set of new, dynamic units of visual speech which are learnt using computer vision and machine learning techniques. Rather than clustering phoneme labels as is done traditionally, the visible articulators of a speaker are tracked and automatically segmented into short, visually intuitive speech gestures based on the dynamics of the articulators. The segmented gestures are clustered into dynamic visemes, such that movements relating to the same visual function appear within the same cluster. Speech animation can then be generated on any facial model by mapping a phoneme sequence to a sequence of dynamic visemes, and stitching together an example of each viseme in the sequence. Dynamic visemes model coarticulation and maintain the dynamics of the original speech, so simple blending at the concatenation boundaries ensures a smooth transition. The efficacy of dynamic visemes for computer animation is formally evaluated both objectively and subjectively, and compared with traditional phoneme to static lip-pose interpolation.

This thesis is about improving machine lip-reading, that is, the classification of speech from only visual cues of a speaker. Machine lip-reading is a niche research problem in both areas of speech processing and computer vision. Current challenges for machine lip-reading fall into two groups: the content of the video, such as the rate at which a person is speaking or; the parameters of the video recording for example, the video resolution. We begin our work with a literature review to understand the restrictions current technology limits machine lip-reading recognition and conduct an experiment into resolution affects. We show that high definition video is not needed to successfully lip-read with a computer. The term 'viseme' is used in machine lip-reading to represent a visual cue or gesture which corresponds to a subgroup of phonemes where the phonemes are indistinguishable in the visual speech signal. Whilst a viseme is yet to be formally defined, we use the common working definition: 'a viseme is a group of phonemes with identical appearance on the lips'. A phoneme is the smallest acoustic unit a human can utter. Because there are more phonemes per viseme, mapping between the units creates a many-to-one relationship. Many mappings have been presented, and we conduct an experiment to determine which mapping produces the most accurate classification. Our results show Lee's [82] is best. Lee's classification also outperforms machine lip-reading systems which use the popular Fisher [48] phoneme-to-viseme map. Further to this, we propose three methods of deriving speaker-dependent phoneme-to-viseme maps and compare our new approaches to Lee's. Our results show the sensitivity of phoneme clustering and we use our new knowledge for our first suggested augmentation to the conventional lip-reading system. Speaker independence in machine lip-reading classification is another unsolved obstacle. It has been observed, in the visual domain, that classifiers need training on the test subject to achieve the best classification. Thus machine lip-reading is highly dependent upon the speaker. Speaker independence is the opposite of this, or in other words, is the classification of a speaker not present in the classifier's training data. We investigate the dependence of phoneme-to-viseme maps between speakers. Our results show there is not a high variability of visual cues, but there is high variability in trajectory between visual cues of an individual speaker with the same ground truth. This implies a dependency upon the number of visemes within each set for each individual. Finally, we investigate how many visemes is the optimum number within a set. We show the phoneme-to-viseme maps in literature rarely have enough visemes and the optimal number, which varies by speaker, ranges from 11 to 35. The last difficulty we address is decoding from visemes back to phonemes and into words. Traditionally this is completed using a language model. The language model unit is either: the same as the classifier, e.g. visemes or phonemes; or the language model unit is words. In a novel approach we use these optimum range viseme sets within hierarchical training of phoneme labelled classifiers. This new method of classifier training demonstrates significant increase in classification with a word language network.

Visemes are groups of phonemes that are considered visually indistinguishable, but there are many doubts regarding their use in visual speech recognition. In this thesis, a visual speech recogniser is constructed to test the validity of visemes. Examining the phoneme output of the recogniser shows that it is not possible to construct a viseme grouping that exhibits the required phoneme confusion characteristics. This thesis proves that phonemes, not visemes, are the basic visual unit of speech.

The aim of this work is to improve the naturalness of visual speech synthesis produced automatically from a linguistic input over existing methods. Firstly, the most important contribution is on the investigation of the most suitable speech units for the visual speech synthesis. We propose the use of dynamic visemes instead of phonemes or static visemes and found that dynamic visemes can generate better visual speech than either phone or static viseme units. Moreover, best performance is obtained by a combined phoneme-dynamic viseme system. Secondly, we examine the most appropriate model between hidden Markov model (HMM) and different deep learning models that include feedforward and recurrent structures consisting of one-to-one, many-to-one and many-to-many architectures. Results suggested that that frame-by-frame synthesis from deep learning approach outperforms state-based synthesis from HMM approaches and an encoder-decoder many-to-many architecture is better than the one-to-one and many-to-one architectures. Thirdly, we explore the importance of contextual features that include information at varying linguistic levels, from frame level up to the utterance level. Our findings found that frame level information is the most valuable feature, as it is able to avoid discontinuities in the visual feature sequence and produces a smooth and realistic animation output. Fourthly, we found that the two most common objective measures of correlation and root mean square error are not able to indicate realism and naturalness of human perceived quality. We introduce an alternative objective measure and show that the global variance is a better indicator of human perception of quality. Finally, we propose a novel method to convert a given text input and phoneme transcription into a dynamic viseme transcription in the case when a reference dynamic viseme sequence is not available. Subjective preference tests confirmed that our proposed method is able to produce animation, that are statistically indistinguishable from animation produced using reference data.

Magistro baigiamajame darbe analizuojama lietuvių šnekos vizemų vizualizacijos problema, tiriami lietuvių kalbos vizemų vizualizacijos požymiai, galimybės koduoti SAMPA kodais analizė, analizuojami programinės įrangos kūrimo metodai ir algoritmai. Sukurtas algoritmas sprendžiantis lietuvių šnekos vizemų vizualizacijos problemą. Pasirinktas kompiuterinės grafikos tipas, tinkantis iškeltiems tikslams įvykdyti – vektorinė grafika. Sukurti du vektorių transformacijos metodai, išanalizuoti jų skirtumai ir praktinio panaudojimo galimybės. Sukurta programinė įranga įgalinanti vartotoją kurti vizemas, jas transformuoti ir derinti jų vaizdavimo trukmę įvairiais koeficientais ir vykdyti animaciją pagal pasirinktą transformacijos metodą. Sudarytos penkios vizemos ir dviejų lietuviškų žodžių animacijos, kurių pagalba atliktas tyrimas parodantis darbo ir metodų realizavimo kokybę bei pritaikomumą.
In this final Master's thesis work features of Lithuanian speech visemes visualization are analyzed. Possibility of coding with SAMPA codes, software methods and algorithms are inspected. Type of computer graphics is picked, which is suitable for software objectives – vector graphics. Two transformation methods for vector graphics are created and their differences and practical usability are analyzed. Software for visemes creation, transformation and tuning of their duration and duration of transformation between visemes is created and described. The main purpose of this software is to animate Lithuanian speech by the method selected. Five visemes for two Lithuanian words animation is created. Using these visemes research has been done which is showing the quality of realization and adaptability of this software.

The goal of this project is to implement a system to analyse an audio signal containing speech, and produce a classifcation of lip shape categories (visemes) in order to synchronize the lips of a computer generated face with the speech.

The thesis describes the work to derive a method that maps speech to lip move- ments, on an animated face model, in real time. The method is implemented in C++ on the PC/Windows platform. The program reads speech from pre-recorded audio files and continuously performs spectral analysis of the speech. Neural networks are used to classify the speech into a sequence of phonemes, and the corresponding visemes are shown on the screen.

Some time delay between input speech and the visualization could not be avoided, but the overall visual impression is that sound and animation are synchronized.

Automatic segmentation of speech is compulsory for building large speech databases to be used in speech processing applications. This study proposes a bimodal automatic speech segmentation system that uses either articulator motion information (AMI) or visual information obtained by a camera in collaboration with auditory information. The presence of visual modality is shown to be very beneficial in speech recognition applications, improving the performance and noise robustness of those systems. In this dissertation a significant increase in the performance of the automatic speech segmentation system is achieved by using a bimodal approach. Automatic speech segmentation systems have a tradeoff between precision and resulting number of gross errors. Boundary refinement techniques are used in order to increase precision of these systems without decreasing the system performance. Two novel boundary refinement techniques are proposed in this thesis
a hidden Markov model (HMM) based fine tuning system and an inverse filtering based fine tuning system. The segment boundaries obtained by the bimodal speech segmentation system are improved further by using these techniques. To fulfill these goals, a complete two-stage automatic speech segmentation system is produced and tested in two different databases. A phonetically rich Turkish audiovisual speech database, that contains acoustic data and camera recordings of 1600 Turkish sentences uttered by a male speaker, is build from scratch in order to be used in the experiments. The visual features of the recordings are extracted and manual phonetic alignment of the database is done to be used as a ground truth for the performance tests of the automatic speech segmentation systems.

Es presenta un esquema únic per a la síntesi i anàlisi audiovisual personalitzable realista de seqüències audiovisuals de cares parlants i seqüències visuals de llengua de signes en àmbit domèstic. En el primer cas, amb animació totalment sincronitzada a través d'una font de text o veu; en el segon, utilitzant la tècnica de lletrejar paraules mitjançant la ma. Les seves possibilitats de personalització faciliten la creació de seqüències audiovisuals per part d'usuaris no experts. Les aplicacions possibles d'aquest esquema de síntesis comprenen des de la creació de personatges virtuals realistes per interacció natural o vídeo jocs fins vídeo conferència des de molt baix ample de banda i telefonia visual per a les persones amb problemes d'oïda, passant per oferir ajuda a la pronunciació i la comunicació a aquest mateix col·lectiu. El sistema permet processar seqüències llargues amb un consum de recursos molt reduït, sobre tot, en el referent a l'emmagatzematge, gràcies al desenvolupament d'un nou procediment de càlcul incremental per a la descomposició en valors singulars amb actualització de la informació mitja. Aquest procediment es complementa amb altres tres: el decremental, el de partició i el de composició.
Se presenta un esquema único para la síntesis y análisis audiovisual personalizable realista de secuencias audiovisuales de caras parlantes y secuencias visuales de lengua de signos en entorno doméstico. En el primer caso, con animación totalmente sincronizada a través de una fuente de texto o voz; en el segundo, utilizando la técnica de deletreo de palabras mediante la mano. Sus posibilidades de personalización facilitan la creación de secuencias audiovisuales por parte de usuarios no expertos. Las aplicaciones posibles de este esquema de síntesis comprenden desde la creación de personajes virtuales realistas para interacción natural o vídeo juegos hasta vídeo conferencia de muy bajo ancho de banda y telefonía visual para las personas con problemas de oído, pasando por ofrecer ayuda en la pronunciación y la comunicación a este mismo colectivo. El sistema permite procesar secuencias largas con un consumo de recursos muy reducido gracias al desarrollo de un nuevo procedimiento de cálculo incremental para la descomposición en valores singulares con actualización de la información media.
A shared framework for realistic and personalizable audiovisual synthesis and analysis of audiovisual sequences of talking heads and visual sequences of sign language is presented in a domestic environment. The former has full synchronized animation using a text or auditory source of information; the latter consists in finger spelling. Their personalization capabilities ease the creation of audiovisual sequences by non expert users. The applications range from realistic virtual avatars for natural interaction or videogames to low bandwidth videoconference and visual telephony for the hard of hearing, including help to speech therapists. Long sequences can be processed with reduced resources, specially storing ones. This is allowed thanks to the proposed scheme for the incremental singular value decomposition with mean preservation. This scheme is complemented with another three: the decremental, the split and the composed ones.

Face-to-face dialogue is the most natural mode of communication between humans. The combination of human visual perception of expression and perception in changes in intonation provides semantic information that communicates idea, feelings and concepts. The realistic modelling of speech movements, through automatic facial animation, and maintaining audio-visual coherence is still a challenge in both the computer graphics and film industry.

The work presented in this thesis forms part of a research project which attempts to generate a visualisation of a speaker's mouth from purely acoustic speech signals. The aim is to provide an aid for partially hearing impaired people in which visual information is presented alongside limited acoustic signals, facilitating easier use of the telephone. The system is essentially a low-level speech recogniser in which phonemic information is extracted from the speech waveform and mapped onto visemes generated on a synthetic facial image. This thesis presents a description of a major part of this project, that is, the development of an accurate phoneme discriminator which is capable of speaker independent operation, on continuous speech. The recognition process is realised in three stages: a pre-processor to convert the speech into a suitable parametric form; a pattern recogniser to identify the possible phoneme classes and a post-processor to produce the viseme information. The pattern recognition stage uses a self-organising Kohonen network, followed by a Learning Vector Quantiser (LVQ) to further improve the recognition accuracy. The performance of this stage is highly dependent on the choice of pre-processor used at the input to the network and it is the design of the pre-processor stage that forms a significant part of this work. A novel technique known as the pseudo-cepstrum forms the basis of this pre-processor. Extensive investigations have been conducted into the dependence of performance on a range of parameters, both at the pre-processor stage and within the Kohonen classifier. In particular, a performance comparison of several preprocessor techniques, including the pseudo-cepstrum, has been carried out. Factors affecting both the training and operation of the classifier are also described here, with the sensitivity of recognition performance to the input data, being a major issue. Overall recognition accuracies of 80% have been achieved.

La stabilisation inertielle de ligne de visée est essentiellement un problème de rejet de perturbations : il faut rendre la ligne de visée de la caméra embarquée dans le viseur insensible aux mouvements du porteur. Les méthodes de commande robuste du type H-infini sont bien adaptées à la résolution de ce type de problème, et plus particulièrement l'approche Loop-Shaping qui repose sur des concepts de réglage de l'automatique fréquentielle classique. Cependant, les correcteurs obtenus via cette approche sont généralement d'ordre élevé et donc difficilement implémentables sur le calculateur embarqué du viseur.Dans cette thèse, nous avons proposé des méthodologies de synthèse de correcteurs robustes d'ordre réduit et/ou de structure fixée. Pour cela, nos travaux ont été axés sur :- L'optimisation pour la synthèse H-infini à ordre et/ou structure fixée. Tout d'abord nous avons exploré les possibilités offertes par l'optimisation sous contraintes LMI (Linear Matrix Inequalities). Celles-ci se sont avérées limitées, bien que de nombreux algorithmes aient été proposés dans ce cadre depuis le début des années 90. Ensuite, nous avons opté pour l'optimisation non lisse. En effet des outils numériques récemment développés rendent accessible cette approche, et leur efficacité s'est avéré indéniable.- L'adaptation au cadre particulier du critère H-infini Loop-Shaping.La structure particulière de ce critère de synthèse a été exploitée afin de mieux prendre en compte les pondérations, et d'améliorer la réduction d'ordre du correcteur final. Enfin, une approche basée uniquement sur le réglage graphique d'un gabarit de gain fréquentiel en boucle ouverte est proposée. Ces différentes méthodologies sont illustrées, tout au long de la thèse, sur un viseur dont le modèle a été identifié à partir de mesures expérimentales.

Cette thèse porte sur la commande H∞ par loop-shaping pour les systèmes linéaires à temps invariant d'ordre faible avec ou sans retard et dépendant de paramètres inconnus. L'objectif est d'obtenir des correcteurs H∞ paramétriques, c'est-à-dire dépendant explicitement des paramètres inconnus, pour application à des viseurs gyrostabilisés.L'existence de ces paramètres inconnus ne permet plus l'utilisation des techniques numériques classiques pour la résolution du problème H∞ par loop-shaping. Nous avons alors développé une nouvelle méthodologie permettant de traiter les systèmes linéaires de dimension finie grâce à l'utilissation de techniques modernes de calcul formel dédiées à la résolution des systèmes polynomiaux (bases de Gröbner, variétés discriminantes, etc.).Une telle approche présente de multiples avantages: étude de sensibilités du critère H∞ par rapport aux paramètres, identification de valeurs de paramètres singulières ou remarquables, conception de correcteurs explicites optimaux/robustes, certification numérique des calculs, etc. De plus, nous montrons que cette approche peut s'étendre à une classe de systèmes à retard.Plus généralement, cette thèse s'appuie sur une étude symbolique des équations de Riccati algébriques. Les méthodologies génériques développées ici peuvent s'étendre à de nombreux problèmes de l'automatique, notamment la commande LQG, le filtrage de Kalman ou invariant
This PhD thesis deals with the H∞ loop-shaping design for low order linear time invariant systems depending on unknown parameters. The objective of the PhD thesis is to obtain parametric H∞ controllers, i.e. controllers which depend explicitly on the unknown model parameters, and to apply them to the stabilization of gyrostabilized sights.Due to the unknown parameters, no numerical algorithm can solve the robust control problem. Using modern symbolic techniques dedicated to the solving of polynomial systems (Gröbner bases, discriminant varieties, etc.), we develop a new methodology to solve this problem for finite-dimensional linear systems.This approach shows several advantages : we can study the sensibilities of the H∞ criterion to the parameter variations, identify singular or remarquable values of the parameters, compute controllers which depend explicitly on the parameters, certify the numerical computations, etc. Furthermore, we show that this approach can be extended to a class of linear time-delay systems.More generally, this PhD thesis develops an algebraic approach for the study of algebraic Riccati equations. Thus, the methodology obtained can be extended to many different problems such as LQG control and Kalman or invariant filtering

Thèse (Ph. D.)--Université de Montréal, 2003.
"NQ-82709." "Thèse présentée à la faculté des études supérieures en vue de l'obtention du grade de Ph. D. en sciences de l'éducation, option Fondements de l'éducation." Version électronique également disponible sur Internet.

Orientadores: Leo Pini Magalhães, Fabio Violaro
Tese (doutorado) - Universidade Estadual de Campinas, Faculdade de Engenharia Eletrica e de Computação
Made available in DSpace on 2018-08-04T21:15:42Z (GMT). No. of bitstreams: 1 DeMartino_JoseMario_D.pdf: 2252476 bytes, checksum: 105696f772b91816f761b9d3534cffc1 (MD5) Previous issue date: 2005
Resumo: aparência de uma face sintética ao longo do tempo. Animação facial sincronizada com a fala está relacionada ao controle da movimentação da face virtual comandada pelos eventos fonéticos de uma locução. Tal controle implica na manipulação da face virtual de forma coordenada e em sincro-nismo com o sinal acústico da fala. A coordenação é a1cançada pela reprodução, na face virtual, da movimentação articulatória visível necessária à produção dos sons da fala. O objetivo do trabalho é estudar e propor uma metodologia para a definição de representações para os padrões visuais de movimentação articulatória observáveis na face durante a fala, os denominados visemas. A metodo-logia proposta estabelece visemas dependentes do contexto fonético que contemplam o fenômeno da coarticulação perseveratória e antecipatória. Além disso, a partir da descrição geométrica e temporal de visemas estabelecidos pela análise de um corpus lingüístico do português do Brasil, são derivados modelos para a movimentação da articulação temporomandibular e do tecido dos lábios. Apesar do material fonético utilizado no trabalho estar restrito ao português do Brasil, a metodologia proposta é aplicável a outras línguas
Abstract: Computer facial animation refers to the techniques for specifying and controlling the positioning, motion, and appearance of a synthetic face over time. Speech synchronized facial animation addres-ses the control of a virtual face conducted by the phonetic events of an utterance. Such control implies the manipulation of the virtual face synchronized and coordinated with the speech signal. The coor-dination is achieved by reproducing on the virtual face the visible articulatory movements necessary for speech production. The objective of the work is to study and propose a methodology to establish representations for, the visual articulatory pattems displayed on the face during speech production, the so called visemes. The proposed methodology identifies phonetic context dependent visemes that cope with persevera tive and anticipatory coarticulation. Additionally, the movements of the temporo-mandibular joint and the lip tissue are modelled from a set of visemes established by the analysis of a Brazilian Portuguese linguistic corpus. Although the corpus is restricted to Brazilian Portuguese, the methodology is general enough to be applied to other languages
Doutorado
Engenharia de Computação
Doutor em Engenharia Elétrica

Orientador: Jose Mario De Martino
Dissertação (mestrado) - Universidade Estadual de Campinas, Faculdade de Engenharia Eletrica e de Computação
Made available in DSpace on 2018-08-14T03:40:13Z (GMT). No. of bitstreams: 1 Costa_PaulaDornhoferParo_M.pdf: 1768234 bytes, checksum: 4f592fb36058804db54ffed874fc1ca3 (MD5) Previous issue date: 2009
Resumo: A animação facial por computador sincronizada com a fala permite a implementação de cabeças virtuais que podem contribuir para tornar interfaces humano-computador mais eficientes e atraentes. O presente trabalho apresenta um método de síntese de animação facial 2D baseado em imagens cujo desenvolvimento foi guiado por dois objetivos principais: a reprodução realista da movimentação articulatória visível da fala, incluindo os efeitos da coarticulação, e a possibilidade de implementação do método mesmo em plataformas com capacidades limitadas de processamento e memória, tais como celulares e assistentes pessoais digitais. O método desenvolvido baseia-se em uma base de imagens de visemas dependentes de contexto para o Português do Brasil e adota a técnica de metamorfose entre visemas para a síntese da animação facial. A abordagem proposta representa uma estratégia de síntese alternativa e inovadora, capaz de reproduzir a movimentação articulatória visível da fala, incluindo os efeitos da coarticulação, a partir de uma base de apenas 34 imagens. O trabalho inclui a implementação de um sistema piloto integrado a conversor texto-fala. Adicionalmente, o método de síntese proposto é avaliado através de teste de inteligibilidade da fala. Os resultados desta avaliação indicam que a informação visual fornecida pelas animações geradas pelo sistema contribui para a inteligibilidade da fala em condições de áudio contaminado por ruído. Apesar do trabalho estar restrito ao Português do Brasil, a solução apresentada é aplicável a outras línguas. Palavras-chave: Computação Gráfica, Animação Facial, Visemas, Coarticulação, Metamorfose
Abstract: Speech synchronized facial animation allows the implementation of talking heads that potentially can improve human-computer interfaces making them more efficient and attractive. This work presentsan image based 2D facial animation synthesis method whose development was guided by two main goals: the realistic reproduction of visible speech articulatory movements, including coarticulation effects, and the possibility to implement the method also on limited processing and memory platforms, like mobile phones or personal digital assistants. The developed method is based on an image database of Brazilian Portuguese context dependent visemes and uses the morphing between visemes strategy as facial animation synthesis technique. The proposed approach represents an alternative and innovative synthesis strategy, capable of reproducing the visible speech articulatory movements, including coarticulation effects, from an image database of just 34 images. This work includes the implementation of a pilot system integrated to a text-to-speech synthesizer. Additionally, the proposed synthesis method is evaluated through a speech intelligibility test. The test results indicate that the animations generated by the system contribute to improve speech intelligibility when audio is degraded by noise. Despite the fact this work is restricted to Brazilian Portuguese, the presented solution is applicable to other languages. Keywords: Computer Graphics, Facial Animation, Visemes, Coarticulation, Morphing
Mestrado
Engenharia de Computação
Mestre em Engenharia Elétrica

La melatonine est une neurohormone. Son role comme mediateur de la photoperiode pour la regulation de divers processus physiologiques saisonniers est bien connu, cependant chez l'homme sa fonction exacte n'a pas ete clairement elucidee. Les recents travaux laissent supposer que cette hormone interviendrait dans de nombreux axes therapeutiques tels que : les desordres du sommeil, desordres saisonniers et la depression. Cependant, deux problemes limitent son utilite : sa tres courte demi de vie plasmatique et son manque de selectivite. Afin de surmonter ces difficultes, nous avons entrepris la syntheses de differents heterocycles oxygenes analogues de cette hormone comme les ligands benzodioxaniques, benzodioxiniques et autres chromatiques et chromeniques. Ces molecules ont montre une affinite prometteuse et certaines ont une tres bonne selectivite pour les sous-recepteurs ml#1#a et ml#1#b de l'humain.

La detection de cibles dans l'atmosphere au moyen d'observations satellitales exige une bonne connaissance du fond atmospherique, afin d'estimer au mieux le contraste entre la cible et le fond. La bande v 3 du co 2 a 4,3 m se prete particulierement bien a ce type d'etude. En effet, le gaz carbonique, principal absorbant autour de 4,3 m, a une repartition homogene et connue. Sa concentration n'induit donc pas de fluctuation du rayonnement. Compte tenu de la forte opacite de l'atmosphere autour de cette region spectrale, permettant de s'affranchir des fortes fluctuations d'origines tropospheriques, le fond radiatif ainsi mesure est peu confus. Certains phenomenes atmospheriques peuvent cependant causer des fluctuations du rayonnement a differentes echelles spatiales et/ou temporelles. Ce travail de these permet de les quantifier. Au cours de notre etude, l'effet des nuages, ondes de gravite, aurores, et sprites, sur la luminance de la bande v 3 du co 2, est modelise pour plusieurs geometries d'observation proches du nadir. Pour cela, nous utilisons les codes de transfert radiatif fascode-3, modtran-4 et sharg-4, ce dernier permettant la prise en compte des effets de netl, importants vers 4,3 m. On s'interesse notamment au niveau maximum de la variation de luminance, ainsi qu'a sa variation spatiale. Nous confrontons ensuite les resultats de la modelisation aux donnees a 4,3 m des instruments hirs et spirit a bord des satellites noaa et msx.

Depuis l'apparition du sida et l'identification du vih, la recherche d'anti-retroviraux orientes contre les differentes etapes du cycle de replication du virus s'est particulierement developpee avec plus ou moins de succes. Parmi les differentes possibilites qui s'ouvrent aux therapeutes, la transcriptase inverse (enzyme virale) est consideree comme une des cibles privilegiees des antiviraux. Parallelement, des travaux intensifs sont effectues pour inhiber d'autres virus type cmv, herpes, papillomavirus,. . . , responsables d'infections opportunistes qui accompagnent souvent de ce syndrome. C'est dans ce contexte que nous avons effectue la synthese de nouveaux analogues carbocycliques de nucleosides, famille consideree comme tres prometteuse sur un plan antiviral. Dans un premier temps, nous avons effectue une mise au point recente et originale de la chimie des analogues carbocycliques de nucleosides, actualisee jusqu'en mars 1993. L'obtention de nouveaux carbocycles a ete realisee par la fonctionnalisation d'un epoxyde anti, le (3,4-epoxycyclopentyl)-methyl-tertbutyldimethylsilylether. Ce compose provient de la protection et de l'epoxydation du 1-hydroxymethyl-3-cyclopentene. A partir de ce synthon cle, nous avons decrit deux methodes d'obtention de carbocycles 3-deoxyribo nucleosides qui sont stereoselectives et plus rapides que celles mentionnees dans la litterature: 1) par construction d'une base heterocyclique autour d'une fonction amine, 2) par condensation de la base heterocyclique sur l'epoxyde anti. Cette reaction s'effectue en catalyse basique (nah) ou acide (bf#3. Et#2o, zncl#2. . . ). Ainsi, les analogues carbocycliques de nucleosides originaux suivants ont ete synthetises et caracterises: a) ()-c-3-deoxyribol (2h,6h-5-amino) pyrimidine, b) ()-c-3-deoxyribo-6-chloro-purine, obtenu par deux voies differentes, c) ()-c-3-deoxyribo-6-purine, d) ()-c-3-deoxyribo-thymine. Ces composes possedent peu ou pas d'activite anti-vih. Neanmoins, l'evaluation biologique des produits synthetises sur d'autres virus (cmv, hsv, flu, vzv) a montre en particulier qu'une des molecules synthetisees possede une activite biologique appreciable contre des virus a herpes

Kalbos animacija plačiai naudojama technikos įrenginiuose siekiant kurtiesiems, vaikams, vidutinio ir vyresnio amžiaus žmonėms sudaryti vienodas bendravimo galimybes. Žmonės yra labai jautrūs veido išvaizdos pokyčiams, todėl kalbos animavimas yra sudėtingas procesas, kurio metu žmogaus kalboje atpažinta akustinė informacija (fonemos) yra vizualizuojama naudojant specialiai sumodeliuotas veido išraiškas vadinamas vizemomis. Didžiausią įtaką kalbos animacijos tikroviškumui turi teisingas fonemas atitinkančių vizemų identifikavimas, modeliavimas ir jų išrikiavimas laiko juostoje. Tačiau, norint užtikrinti kalbos animacijos natūralumą, būtina papildomai išnalizuoti vizemų įtaką kaimyninėms fonemoms ir atsižvelgiant į animuojamos kalbos fonetines savybes sukurti koartikuliacijos valdymo modelį. Kiekvienos kalbos fonetika skiriasi, todėl kitai vienai kalbai sukurta animavimo sistema nėra tiesiogiai tinkama kitai kalbai animuoti. Kalbos animavimo karkasas, kuriame realizuojama Lietuvių kalbai skirta animavimo technologija, turi būti sukurta lietuvių kalbai vizualizuoti. Darbą sudaro įvadas, trys pagrindiniai skyriai, bendrosios išvados, literatūros sąrašas, publikacijų sąrašas. Pirmame skyriuje Skyriuje analizuojamos pasaulyje naudojamos kalbos animavimo technologijos. Kalbos signalas yra ir girdimas, ir matomas, todėl jos animacija yra sudėtinis procesas priklausantis nuo pasirinktos veido modeliavimo metodikos, kalbos signalo tipo, ir koartikuliacijos valdymo modelio. Antrajame... [toliau žr. visą tekstą]
Speech animation is widely used in technical devices to allow the growing number of hearing impaired persons, children, middle-aged and elderly equal participation in communication. Speech animation systems (“Talking heads”) are basically driven by speech phonetics and their visual representation – visemes. Acuraccy of the chosen speech recognition engine, naturally looking visemes, phoneme to viseme mapping and coarticulation control model considerably influence the quality of animated speech. Speech animation is strongly related with language phonetics, so new“Talking heads” should be created to animate different languages. Framework suitable to animate Lithuanian speech, which includes two new models that help to improve intelligibility of animated Lithuanian speech is used to create Lithuanian „Talking head” „LIT”. The dissertation consists of Introduction, three main chapters and general conclusions. Chapter 1 provides the analysis of the existing speech animation technologies. Different facial modelling techniques are analysed to define the most suitable 3D „Talking head” modelling technique for Lithuanian language. Viseme classification experiments across different languages are analysed to identify variety of viseme classification methods. Coarticulation control models are compared to deside which one should be used to define coarticulation of Lithuanian speech. Chapter 2 describes theoretical framework for Lithuanian speech animation. Translingual visual speech... [to full text]

Systemes d'hyperthermie capacitive radiofrequences a 2 et 3 electrodes (13 mhz). Systemes d'hyperthermie radiative micro-onde (2450,915 mhz). Mesure des distributions de puissance et de temperance dans differentes conditions de chauffage hyperthermique utilisait des fantomes equivalent-tissus biologiques et des dispositifs de thermometrie non interferents avec les ondes electromagnetiques

Cette these etudie la personnalite de isaac baker brown, eminent gynecolo, gue londonnien, qui a largement contribue, au cours de la seconde moitie du xixe siecle, aux progres de la gynecologie, discipline naissante a cette epoque. Or, la derniere innovation de ce praticien fut de pratiquer des clitoridectomies envisagees comme traitement de divers troubles psychiatriques. Cette pratique et la publicite qu'il en fit lui valurent d'etre mis en accusation par la societe royale d'obstetrique et d'en etre radie. L'objectif de ce travail, apres avoir decrit et analyse l'exercice de la medecine au xixe siecle est de souligner l'attitude moralisa+ice de la societe victorienne vis-a-vis de la sexualite feminine. De plus il doit permettre de saisir le role joue par la presse medicale dans la "mise a mort" professionnelle d'un confrere medecin
This thesis makes a study of isaac baker brown's personality, a distinguished london gynaecologist who amply contributed to the improvements of gynaecolo, gy, a nascent discipline, during the second half of the 19th century. Then this doc, tor's last innovation was the practice clitoridectomy as a cure for some psychiatric disorders. This practice and the advertisement he gave to it were the cause of his accusation by the royal college of gynaecology and his eviction. After having described and analysed the medical practice during the 19th century, the purpose of this work is to underline the moralizing attitude of the victo, rian society towards female sexuality. More, it may help to understand the part played by the medical press in the professional downfall of a medical colleague

Esta dissertação de mestrado apresenta um banco de figuras e palavras. O banco objetiva servir para a aumentar a validade e precisão dos instrumentos de avaliação, bem como a eficácia dos materiais instrucionais para desenvolvimento de linguagem. Este banco contém 971 entradas lexicais. Cada entrada contém uma figura e seu correspondente nome escrito. A figura é analisada em termos de grau de univocidade (grau de concordância na nomeação). O nome da figura é analisado separadamente em três formas: palavra escrita visível, palavra falada audível, e palavra falada visível (lida orofacialmente). Palavras escritas visíveis são compostas de grafemas. São analisadas em termos de seu grau de familiaridade e reconhecibilidade (grau em que pode ser lida via rota lexical) e decodificabilidade (grau em que pode ser lida pela rota perilexical ou fonológica). Palavras ouvidas audíveis são compostas de fonemas. São analisadas em termos do grau de cifrabilidade (grau de facilidade com que podem ser escritas via rota perilexical). Palavras faladas vistas são compostas de fanerolaliemas. São analisadas em termos do grau de legibilidade orofacial (grau em que podem ser compreendidas apenas pela visão), audibilizabilidade (grau em que a imagem auditiva dos fonemas pode ser evocada por fanerolaliemas durante a leitura orofacial visual), e cifrabilidade (grau de facilidade com que podem ser escritas via rota perilexical). O banco é composto de 971 entradas lexicais, cada qual composta de uma figura (à esquerda) e de vários dados pertinentes ao seu nome correspondente (à direita). O lado direito da entrada é composto de seis campos. O Campo 1 fornece o o nome da figura escrito em dois alfabetos: alfabeto romano e alfabeto fonético internacional. Ele também fornece a categoria semântica à qual pertence a palavra. O Campo 2 fornece o número da figura (para indexar todas as 971 figuras do banco). O Campo 3 fornece a univocidade da figura numa escala de 0-100 pontos separadamente para crianças de 2, 3, 4, 5, 6 7-10 anos de idade, bem como para adultos. O Campo 4 fornece a palavra escrita visível tal como analisada em termos de seu grau de familiaridade ou reconhecibilidade (grau em que pode ser lida via rota lexical) numa escala de 1-9 pontos, separadamente para crianças de 5º ano, 4º ano, 3º ano, 2º ano, e 1º ano. Nessa escala, 5 corresponde à média, 6 a 1 erro-padrão (EP) acima da média, 7 a 2 EP acima da média e assim por diante até 9; ao passo que 4 corresponde 1 EP abaixo da média, 3 a 2 EP abaixo da média, e assim por diante até 1. O Campo 5 é composto de quatro linhas, cada qual dividida em quatro colunas. A Linha 1 fornece o grau de decifrabilidade (grau com que pode ser lida pela rota perilexical) da palavra escrita visível, numa escala de 0-1. A Linha 2 fornece o grau de cifrabilidade da palavra ouvida (grau com que pode ser escrita pela rota perilexical), numa escala de 0-1. A Linha 3 fornece o grau de audibilizabilidade da palavra falada lida orofacialmente (grau com que sequência de fanerolaliemas pode ser convertida em sequência de fonemas), numa escala de 0-1. A Linha 3 fornece o grau de cifrabilidade da palavra falada lida orofacialmente (grau com que sequência de fanerolaliemas pode ser convertida em sequência de grafemas), numa escala de 0-1. Cada palavra é dividida em suas colunas. cada coluna fornece os dados referentes à linha em questão em uma de quatro formas diferentes. Nas Colunas 1 e 2, dados consistem na média das razões independente da incidência. Nas Colunas 3 e 4, dados consistem na média das razões ponderada por incidência diferencial. Nas Colunas 1 e 3 os dados consistem na média das razões independentemente da tonicidade da fala (seja ouvida ou vista) na pronúncia. Nas Colunas 2 e 4, os dados consistem na média das razões ponderada pela tonicidade diferencial da fala (seja ouvida ou vista) na pronúncia. Por exemplo, a Linha 1 fornece o grau de decifrabilidade grafema-fonema da palavra escrita visível. Na Coluna 1 decoficabilidade é calculada como mera média de razões independente da incidência ou tonicidade. Na Coluna 2 decodificabilidade é calculada como média das razões independente da incidência mas ponderasa pela tonicidade. Na Coluna 3 decodificabilidade é calculada como média de razões ponderadas em termos de incidência mas independente de tonicidade. Na Coluna 4 decodificabilidade é calculada como média de razões ponderadas em termos de incidência e de tonicidade. O Campo 6 fornece o grau de legibilidade orofacial da fala vista, numa escala de 0-1. O grau de legibilidade orofacial é apresentado em quatro formas. Nas Colunas 1 e 2 ela se encontra calculada segundo o modelo Dória; nas Colunas 3 e 4 ela se encontra calculada segundo o modelo Fonético-Articulatório. Nas Colunas 1 e 3 ela é calculada independentemente da tonicidade da pronúncia; nas Colunas 2 e 4 ela é calculada de modo ponderado pela tonicidade diferencial da pronúncia
This master\'s thesis presents a new sourcebook aimed at increasing the validity and precision of language assessment tools, as well as the efficacy of instructional materials for language development. The sourcebook contains 971 lexical entries. Each entry contains a picture and its corresponding written name. The picture is analyzed in terms of its degree of univocity (i.e., picture naming agreement). The picture name is analyzed separately in three forms: visual written word, auditory spoken word, and visual spoken word (i.e., speechreading). Visual written word is made of graphemes. It is analyzed in terms of its degree of both: familiarity or recognizability (i.e., the degree to which it is suitable to be read via lexical reading route) and decodibility (i.e., the degree to which it is suitable to be read via perilexical reading route). Auditory spoken word is made of phonemes. It is analyzed in terms of its degree of encodibility (i.e., the degree to which it may be suitable for writing or spelling via perilexical spelling route). Visual spoken word is made of visemes. It is analyzed in terms of its degree of: speechreadability (i.e., the degree to which it may be understood via visual speechreading), audibilizability (i.e., the degree to which the auditory imagery of phonemes can be evoked by mouthshapes or visemes during speechreading), and encodibility (i.e., the degree to which it is suitable to be written or spelled correctly via perilexical route). The sourcebook is made of 971 lexical entries. Each entry is made of a picture (on the left) and several data pertaining to its corresponding name (on the right). The right side of the entry is made of six areas. The first area provides the picture name as it is written in both alphabets: the Roman alphabet (orthographic form) and the International Phonetic Alphabet. It also provides the semantic category to which the word belongs. The second area provides the picture number (for indexing all the 971 pictures of the sourcebook). The third area provides the picture univocity in a 0-100 scale for children aged: 2 years, 3 years, 4 years, 5 years, 6 years, 7 to 10 years, as well as for adults. The fourth area provides the visual written word as it is analyzed in terms of its degree of familiarity or recognizability (i.e., the degree to which the written word is suitable to be read via lexical reading route) in a 1-9 point scale, for children from 5th grade, 4th grade, 3rd grade, 2nd grade, and 1st grade. In such a scale, 5 corresponds to the mean, 6 is the mean plus 1 standard error, 7 is the mean plus 2 standard errors and so forth until 9, whereas 4 corresponds to the mean minus 1 standard error, 3 corresponds to the mean minus 2 standard errors and so forth until 1, which corresponds to the mean minus 4 standar erros. The fifth area is made of four lines. Each line is divided into four columns. The first line provides the visual written word degree of decodibility (i.e., the degree to which it is suitable to be read via perilexical reading route) in a 0-1 scale. The second line provides the auditory spoken word degree of encodibility (i.e., the degree to which it may be suitable for writing or spelling via perilexical spelling route) in a 0-1 scale. The third line provides the visual spoken word degree of audibilizability (i.e., the degree to which the auditory imagery of phonemes can be evoked by mouthshapes or visemes during speechreading) in a 0-1 scale. The fouth line provides the visual spoken word degree of encodibility (i.e., the degree to which it is suitable to be written or spelled correctly via perilexical route) in a 0-1 scale. Each line is divided into four columns. Each column presents the data pertaining to the line in question in 1 of 4 different forms. In the first and second columns the data consist of the mean of the ratios regardless of incidence. In the third and fourth columns the data consist of the mean of the ratios weighted by differencial incidence. In the first and third columns the data consist of the mean of the ratios regardless of tonicity of speech (either auditory or visual) in pronunciation. In the second and fourth columns the data consist of the mean of the ratios weighted by differencial tonicity of speech (either auditory or visual) in pronunciation. For instance the first line provides the visual written word degree of decodibility (i.e., grapheme to phoneme decoding). In the first column decodibility is calculated as a mere mean of the ratios regardless of either incidence or tonicity. In the second column decodibility is calculated as a mean of the ratios regardless of incidence but weighted in terms of tonicity. In the third column decodibility is calculated as a mean of the ratios weighted in terms of incidence but regardless of tonicity. In the fourth column it is calculated as a mean of the ratios weighted in terms of both incidence and tonicity. The sixth area provides the visual spoken word degree of speechreadability (i.e., the degree to which it may be understood via visual speechreading) in a 0-1 scale. The speechreadability is presented in 1 of 4 different forms. In the first and second columns, the speechreadability is calculated according to Doria\'s model. In the third and fourth columns it is calculated according to a phonetic model. In the first column and third columns it is calculated regardless of tonicity in pronunciation. In the second and fourth columns it is calculated in a way that is weighted by the differencial tonicity in pronunciation

Seules les donnees spatiales permettent en antarctique d'avoir des donnees a l'echelle globale. Mais la neige est un milieu tres complexe et de nombreux parametres de la physique de la mesure satellitaire restent irresolus. Le diffusiometre, instrument concu pour mesurer la direction et l'intensite du vent au-dessus de l'ocean peut permettre de repondre a ces problemes en raison des nombreux parametres qu'il contient. Apres une discussion des resultats deja obtenus au-dessus de l'antarctique par d'autres mesures spatiales (altimetrie et radiometrie microonde) et une description du fonctionnement du diffusiometre, une etude semi-empirique montre qu'un modele d'echo de surface a deux echelles de rugosite semble bien representer les donnees de cet instrument. En effet, les vents catabatiques creent de petites ondulations a l'echelle centimetrique et des dunes de neige d'une dizaine de centimetres allongees dans le sens du vent. Le signal diffusiometrique decroit de 10 a 20 db avec l'incidence. Un signal additionnel de 5 db est observe qui depend de l'azimut de visee. La direction et l'intensite du vent sont ainsi determinees dans le secteur observe. Ces vents sont en excellent accord avec ceux estimes a l'aide d'un modele numerique des ecoulements atmospherique. Ce travail met en evidence l'importance du vent dans l'interpretation des autres donnees spatiales

On desire utiliser les deserts pour l'etalonnage (multi-angulaire, multi-temporel, multi-bande) des capteurs optiques spatiaux. 20 zones sont selectionnees dans toute l'afrique du nord et l'arabie saoudite a l'aide de donnees meteosat (variabilite spatiale inferieure a 3%, pluviometrie faible, couverture nuageuse faible, effets directionnels modelisables, variabilite temporelle inferieure a 2%). La reflectance bidirectionnelle spectrale ou reflectance de reference est modelisee a l'aide de campagnes de mesures pour un nombre de zones limite. Un premier bilan d'erreur de la methode d'etalonnage par visee de zones desertiques est realise a l'aide de donnees avhrr

Le travail est consacre a l'examen de l'activite critique de deux auteurs francais: roalnd barthes et gilbert durand. Le but principal de la these consiste a confronter les attitudes intellectuelles, les methodes d'approche des oeuvres de culture et les resultats finaux de la recherche menee par les deux critiques que, selon une opinion courante, "tout" separe. La premiere partie presente ainsi les "silhouettes" intellectuelles" de r. Barthes et de g. Durand, ce qui permet de determiner les champs communs de leurs investigations dans le domaine des sciences humaines. La deuxieme partie est consacree a l'etude des aspects methodologiques de leurs conceptions et accentue surtout leur effort commun pour depasser l'antithese de la perspective objectiviste de l'approche des oeuvres et leur approche subjectiviste. La troisieme partie rapporte les idees des auteurs concernes a la critique des oeuvres de narration. L'etude des aspects des oeuvres que r. Barthes et g. Durand mettent en relief permet de reflechir sur les bases intellectuelles et imaginaires de leurs conceptions ainsi que sur les rapports qui existeraient entre les choix methodologiques effectues par les critiques et les conceptions du monde plus generales qui emanent de leurs travaux. La quatrieme partie, qui constitue en meme temps une conclusion elargie, se propose de situer les deux auteurs dans la perspective plus large des tendances generales de la culture occidentale au seuil du xxie siecle et presente l'oeuvre de r. Barthes et de g. Durand comme reflechissant l'etat actuel de la reflexion sur le phenomene de lasignification
The thesis is devoted to analysis of the activity of roland barthes and gilbert durand. Its main purpose is to compare and relate the kind of imagination, research methods and the results of findings of these two critics that are generally considered as representing two extremes of humanities. In the first part of the paper the intellectual personnalities of r. Barthes and g. Durand are presented. This allows to indicate the basics elements on which their thoughts are based. The second part is devoted to the methodological aspect of research activity of these two critics as far as writing are concerned. It has been found that they both, maybe in different ways, examine the possibility of reconciling the objective analysis of literary work with the subjective approach, usually considered as being in opposition. The third part relates the former results of research (presented in the first part of the paper) to r. Barthes' and g. Durand's critical reflections, mainly about narrative writing of the 19th and 20 th centuries. On this basis the relationships between the attitudes towards literary works and outlooks on life represented by these two autors are analysed. The fourth part, which is also an extended conclusions, aims at thinking over the place of r. Barthes' and g. Durand's activity on the background of general trends of the culture nowadays

Ce travail decrit la synthese de deux familles d'aminosaccharides a visee therapeutique. La premiere gamme comprend des derives du d-galactose, du d-glucose, du d-xylose et du xylitol, sur lesquels ont ete greffes par l'intermediaire d'une jonction aminee, soit une chaine n-alkyle soit un groupement phenylalkyle. Trois methodes ont ete utilisees, sur le site anomerique ; condensation d'une amine primaire, sur les sites primaires : greffe du nh#3 ou de rnh#2 apres activation, sur les sites secondaires, d'abord preparation d'un aminosaccharide selon la sequence suivante, activation, azidation, reduction puis alkylation ou acylation. Des etudes biologiques in vivo et in vitro ont montre que les derives n-octylamine sont des antagonistes des canaux calciques membranaires. Les composes de la seconde famille sont des derives dont l'un des oh a ete substitue par un groupement piperazinique. A partir du d-glucose, du d-galactose et du d-xylose, la condensation a ete effectuee selectivement sur la position anomerique pour conduire a des derives pyraniques de configuration. A partir du d-glucose et du d-galactose proteges et actives, les substrats piperaziniques ont ete greffes sur la position primaire soit par substitution nucleophile, soit par addition nucleophile. La liberation controlee des hydroxiles s'est effectuee par catalyse acide en milieu homogene. A partir du 5-desoxy-1,2-0-isopropylidene-5-n-octylamino- -d-xylofuranose, la deprotection conduit au sel de 3-hydroxy-1-n-octyloyridium avec un rendement quantitatif. Des premiers resultats montrent que certains de ces derives inhibent la croissance de bacteries pathogenes.

Ce travail traite de la synthese et de l'evaluation biologique de 9-oxaquinolizidines, inhibiteurs potentiels de glycosidases a visee anti-sida. La conception de ces nouveaux inhibiteurs de glycosidases possedant une fonction perhydro-oxazine, a ete realisee en se basant sur le mecanisme d'hydrolyse des glycosides. Le principe des syntheses repose sur l'utilisation d'hexopyranoses comme produit de depart, de meme stereochimie que les composes souhaites. Cette derniere a ete conservee via des reactions d'amination reductrice stereoselectives. Les reactivites observees en serie glucose, mannose et xylose tres differentes des intermediaires, ont necessite l'utilisation de sequences reactionnelles differentes et specifiques. Celles-ci nous ont conduit a developper trois nouvelles reactions: 1) une nouvelle synthese de perhydro-oxazines a partir d'hydroxylactames; 2) une synthese stereoselective de tetrahydrofuranes fonctionnalises en milieu neutre ou basique; 3) une reaction d'inversion au niveau du carbone 2 de lactones derivees du glucose et du mannose, induite par l'iodure de magnesium

Baigiamajame darbe tiriamas lietuvių kalbos garsų ir jų vaizdinės informacijos ryššys. Atliekama kalbančių galvų modelių animavimo algoritmų analizė, išškeliama jų problematika ir atsižvelgiant į tai pasiūloma lietuvių kalbos sintetinimo metodika, kuri yra pagrįsta anglų kalbos vizemų naudojimu. ŠŠiame darbe sukuriama 30 trimačių lietuvių kalbos vizemų, kurias vizualiai lyginant su standartinėmis anglų kalbos fonemų vizemomis, sudaroma lietuvišškų fonemų ir anglišškų vizemų atitikčių lentelė. Sudaryta lentelė naudojama lietuvių kalbos garso rinkmenai animuoti.
The connection of Lithuanian sounds and their visual aspect is analyzed. The thesis consists of talking head animation algorithms analysis, problematic topics. In reference it is proposed the idea, how to synthesize Lithuanian speech using English visemes. 30 three dimensional Lithuanian visemes are created. After visual comparison of 3D Lithuanian and standard English visemes, the table of Lithuanian phonemes and English visemes mapping is created. The table is used for animating the Lithuanian sound file.