In terms of master’s thesis HAM-Tools library designed for MATLAB/Simulink was modified for the use in simulations of houses in the Czech Republic. Modified library and its parts were described in detail and tested by the simulation of the one-zone and two-zones models of the house. The simulations of models with same parameters were also realized in program TRNSYS. The corresponding results achieved in mentioned simulation tools were compared to each other. The one-zone model created by using HAM-Tools library is tested by the simulation of ventilating, heating, cooling, and sources of moisture. A demonstration of the practical use of the simulation is carried out in the thesis, namely by examining the influence of the insulation thickness on the thermal performance of the house (resp. its heat loss) on real atmospheric conditions. Among others, available resources of meteorological data are mentioned and compared to each other. The function for processing of the meteorological data to a file compatible with the HAM-Tools library was created. It was also created a material data file containing commonly used materials of building structures in the Czech Republic and their parameters.

La présente thèse traite de l'analyse de scènes extraites d'environnements sonores, résultat auditif du mélange de sources émettrices distinctes et concomitantes. Ouvrant le champ des sources et des recherches possibles au-delà des domaines plus spécifiques que sont la parole ou la musique, l'environnement sonore est un objet complexe. Son analyse, le processus par lequel le sujet lui donne sens, porte à la fois sur les données perçues et sur le contexte de perception de ces données.Tant dans le domaine de la perception que de l'apprentissage machine, toute expérience suppose un contrôle fin de l'expérimentateur sur les stimuli proposés. Néanmoins, la nature de l'environnement sonore nécessite de se placer dans un cadre écologique, c'est à dire de recourir à des données réelles, enregistrées, plutôt qu'à des stimuli de synthèse. Conscient de cette problématique, nous proposons un modèle permettant de simuler, à partir d'enregistrements de sons isolés, des scènes sonores dont nous maîtrisons les propriétés structurelles -- intensité, densité et diversité des sources. Appuyé sur les connaissances disponibles sur le système auditif humain, le modèle envisage la scène sonore comme un objet composite, une somme de sons sources.Nous investissons à l'aide de cet outil deux champs d'application. Le premier concerne la perception, et la notion d'agrément perçu dans des environnements urbains. L'usage de données simulées nous permet d'apprécier finement l'impact de chaque source sonore sur celui-ci. Le deuxième concerne la détection automatique d'événements sonores et propose une méthodologie d'évaluation des algorithmes mettant à l'épreuve leurs capacités de généralisation. / This thesis deals with environmental scene analysis, the auditory result of mixing separate but concurrent emitting sources. The sound environment is a complex object, which opens the field of possible research beyond the specific areas that are speech or music. For a person to make sense of its sonic environment, the involved process relies on both the perceived data and its context. For each experiment, one must be, as much as possible,in control of the evaluated stimuli, whether the field of investigation is perception or machine learning. Nevertheless, the sound environment needs to be studied in an ecological framework, using real recordings of sounds as stimuli rather than synthetic pure tones. We therefore propose a model of sound scenes allowing us to simulate complex sound environments from isolated sound recordings. The high level structural properties of the simulated scenes -- such as the type of sources, their sound levels or the event density -- are set by the experimenter. Based on knowledge of the human auditory system, the model abstracts the sound environment as a composite object, a sum of soundsources. The usefulness of the proposed model is assessed on two areas of investigation. The first is related to the soundscape perception issue, where the model is used to propose an innovative experimental protocol to study pleasantness perception of urban soundscape. The second tackles the major issue of evaluation in machine listening, for which we consider simulated data in order to powerfully assess the generalization capacities of automatic sound event detection systems.

<p>Breeding for improved animal welfare is paramount for increasing the long-term sustainability of the animal food industry. In this context, the main objectives of this dissertation were to understand the genetic and genomic background of welfare indicators in livestock and evaluate the feasibility of single-step Genomic Best Linear Unbiased Prediction (ssGBLUP) for performing genomic selection in beef cattle. This dissertation includes five studies. First, we aimed to test and identify an optimal ssGBLUP scenario for crossbreeding schemes. We simulated multiple populations differing based on the genetic background of the trait, and then we tested alternative models, such as multiple-trait weighted ssGBLUP. Even though more elaborated scenarios were evaluated, a single-trait ssGBLUP approach was recommended when genetic correlation across populations were higher than 0.70. The goal of the second study was to identify genomic regions controlling behavior traits that are conserved across livestock species. We systematically reviewed genomic regions associated with behavioral indicators in beef and dairy cattle, pigs, and sheep. The genomic regions identified in this study were located in genes previously reported controlling human behavioral, neural, and mental disorders. In the third study we used a large dataset (675,678 records) from North American Angus cattle to investigate the genetic background of temperament, a behavioral indicator, recorded on one-year-old calves, and provide the models and protocols for implementing genomic selection. We reported a heritability estimate equal to 0.38 for yearling temperament, and it was, in general, genetically favorably correlated with other productivity and fertility traits. Candidate genomic regions controlling yearling temperament were also identified. The fourth study was based on temperament recorded on North American Angus cows from 2 to 15 years of age (797,187 records). The goal was to understand the genetic and genomic background of temperament across the animal’s lifetime. By fitting a random regression model, we observed that temperament is highly genetically correlated across time. However, animals have differential learning and behavioral plasticity (LBP; changes of the phenotype overtime), although the LBP heritability is low. In our last study we evaluated foot scores (foot angle, FA; and claw set, CS) in American (US) and Australian (AU) Angus cattle aiming to assess the genetic and genomic background of foot scores and investigate the feasibility of performing an across-country genomic evaluation (~1.15 million animals genotyped). Foot scores are heritable (heritability from 0.22 to 0.27), and genotype-by-environment interaction was observed between US and AU Angus populations (genetic correlation equal to 0.61 for FA and 0.76 for CS). An across-country genomic prediction outperformed within-country evaluations in terms of predictivity ability (bias, dispersion, and validation accuracy) and theoretical accuracies. We have also identified genes associated with FA and CS previously reported in human’s bone structure and repair mechanism. In conclusion, this dissertation presents a comprehensive genetic and genomic characterization of welfare indicators (temperament and foot scores) in (inter)national livestock populations. </p>

L’imagerie par contraste de phase suscite un intérêt croissant dans le domaine biomédical, puisqu’il offre un contraste amélioré par rapport à l’imagerie d’atténuation conventionnelle. En effet, le décalage en phase induit par les tissus mous, dans la gamme d’énergie utilisée en imagerie, est environ mille fois plus important que leur atténuation. Le contraste de phase peut être obtenu, entre autres, en laissant un faisceau de rayons X cohérent se propager librement après avoir traversé un échantillon. Dans ce cas, les signaux obtenus peuvent être modélisés par la diffraction de Fresnel. Le défi de l’imagerie de phase quantitative est de retrouver l’atténuation et l’information de phase de l’objet observé, à partir des motifs diffractés enregistrés à une ou plusieurs distances. Ces deux quantités d’atténuation et de phase, sont entremêlées de manière non-linéaire dans le signal acquis. Dans ces travaux, nous considérons les développements et les applications de la micro- et nanotomographie de phase. D’abord, nous nous sommes intéressés à la reconstruction quantitative de biomatériaux à partir d’une acquisition multi-distance. L’estimation de la phase a été effectuée via une approche mixte, basée sur la linéarisation du modèle de contraste. Elle a été suivie d’une étape de reconstruction tomographique. Nous avons automatisé le processus de reconstruction de phase, permettant ainsi l’analyse d’un grand nombre d’échantillons. Cette méthode a été utilisée pour étudier l’influence de différentes cellules osseuses sur la croissance de l’os. Ensuite, des échantillons d’os humains ont été observés en nanotomographie de phase. Nous avons montré le potentiel d’une telle technique sur l’observation et l’analyse du réseau lacuno-canaliculaire de l’os. Nous avons appliqué des outils existants pour caractériser de manière plus approfondie la minéralisation et les l’orientation des fibres de collagènes de certains échantillons. L’estimation de phase, est, néanmoins, un problème inverse mal posé. Il n’existe pas de méthode de reconstruction générale. Les méthodes existantes sont soit sensibles au bruit basse fréquence, soit exigent des conditions strictes sur l’objet observé. Ainsi, nous considérons le problème inverse joint, qui combine l’estimation de phase et la reconstruction tomographique en une seule étape. Nous avons proposé des algorithmes itératifs innovants qui couplent ces deux étapes dans une seule boucle régularisée. Nous avons considéré un modèle de contraste linéarisé, couplé à un algorithme algébrique de reconstruction tomographique. Ces algorithmes sont testés sur des données simulées. / Phase contrast imaging has been of growing interest in the biomedical field, since it provides an enhanced contrast compared to attenuation-based imaging. Actually, the phase shift of the incoming X-ray beam induced by an object can be up to three orders of magnitude higher than its attenuation, particularly for soft tissues in the imaging energy range. Phase contrast can be, among others existing techniques, achieved by letting a coherent X-ray beam freely propagate after the sample. In this case, the obtained and recorded signals can be modeled as Fresnel diffraction patterns. The challenge of quantitative phase imaging is to retrieve, from these diffraction patterns, both the attenuation and the phase information of the imaged object, quantities that are non-linearly entangled in the recorded signal. In this work we consider developments and applications of X-ray phase micro and nano-CT. First, we investigated the reconstruction of seeded bone scaffolds using sed multiple distance phase acquisitions. Phase retrieval is here performed using the mixed approach, based on a linearization of the contrast model, and followed by filtered-back projection. We implemented an automatic version of the phase reconstruction process, to allow for the reconstruction of large sets of samples. The method was applied to bone scaffold data in order to study the influence of different bone cells cultures on bone formation. Then, human bone samples were imaged using phase nano-CT, and the potential of phase nano-imaging to analyze the morphology of the lacuno-canalicular network is shown. We applied existing tools to further characterize the mineralization and the collagen orientation of these samples. Phase retrieval, however, is an ill-posed inverse problem. A general reconstruction method does not exist. Existing methods are either sensitive to low frequency noise, or put stringent requirements on the imaged object. Therefore, we considered the joint inverse problem of combining both phase retrieval and tomographic reconstruction. We proposed an innovative algorithm for this problem, which combines phase retrieval and tomographic reconstruction into a single iterative regularized loop, where a linear phase contrast model is coupled with an algebraic tomographic reconstruction algorithm. This algorithm is applied to numerical simulated data.