It is commonly assumed that the goal of stereovision is computing explicit 3D scene reconstructions. We show that very accurate camera calibration is needed to support this, and that such accurate calibration is difficult to achieve and maintain. We argue that for tasks like recognition, figure/ground separation is more important than 3D depth reconstruction, and demonstrate a stereo algorithm that supports figure/ground separation without 3D reconstruction.

This thesis focuses on applications of classical tools from probability theory and convex analysis such as limit theorems to problems in theoretical computer science, specifically to pseudorandomness and learning theory. At first look, limit theorems, pseudorandomness and learning theory appear to be disparate subjects. However, as it has now become apparent, there's a strong connection between these questions through a third more abstract question: what do random objects look like. This connection is best illustrated by the study of the spectrum of Boolean functions which directly or indirectly played an important role in a plethora of results in complexity theory. The current thesis aims to take this program further by drawing on a variety of fundamental tools, both classical and new, in probability theory and analytic geometry. Our research contributions broadly fall into three categories. Probability Theory: The central limit theorem is one of the most important results in all of probability and richly studied topic. Motivated by questions in pseudorandomness and learning theory we obtain two new limit theorems or invariance principles. The proofs of these new results in probability, of interest on their own, have a computer science flavor and fall under the niche category of techniques from theoretical computer science with applications in pure mathematics. Pseudorandomness: Derandomizing natural complexity classes is a fundamental problem in complexity theory, with several applications outside complexity theory. Our work addresses such derandomization questions for natural and basic geometric concept classes such as halfspaces, polynomial threshold functions (PTFs) and polytopes. We develop a reasonably generic framework for obtaining pseudorandom generators (PRGs) from invariance principles and suitably apply the framework to old and new invariance principles to obtain the best known PRGs for these complexity classes. Learning Theory: Learning theory aims to understand what functions can be learned efficiently from examples. As developed in the seminal work of Linial, Mansour and Nisan (1994) and strengthened by several follow-up works, we now know strong connections between learning a class of functions and how sensitive to noise, as quantified by average sensitivity and noise sensitivity, the functions are. Besides their applications in learning, bounding the average and noise sensitivity has applications in hardness of approximation, voting theory, quantum computing and more. Here we address the question of bounding the sensitivity of polynomial threshold functions and intersections of halfspaces and obtain the best known results for these concept classes.

It is widely acknowledged that elevated levels of noise are commonplace in the healthcare environment, particularly in high acuity areas such as the operating room (OR). Excessive ambient noise may pose a threat to patient safety by adversely impacting provider performance and interfering with communication among perioperative care team members. With respect to the certified registered nurse anesthetist (CRNA), increased ambient OR noise may engender distractibility, diminish situation awareness and cause untoward health effects, thereby increasing the possibility for the occurrence of error and patient injury. This research project analytically examines the perceived impact of ambient noise in the operating room by CRNAs. Findings from this study reveal that CRNAs perceive elevated noise to be regularly present in the OR, specifically during the critical emergence phase of the anesthetic. However, CRNAs feel that increased noise only occasionally limits their ability to perform procedures, concentrate and communicate with the perioperative team. OR noise rarely interferes with memory retrieval. CRNAs perceive that noise is sometimes a threat to patient safety but infrequently engenders adverse patient outcomes. CRNAs do not perceive noise in the OR to be detrimental to their health but strongly agree that excessive noise can and should be controlled. Increased ambient OR noise is a veritable reality that may pose a potential threat to patient safety. Further research to identify elevations in noise during critical phases of the anesthetic and delineation of significant contributors to its genesis is warranted.

Die hohe Inzidenz von 300 bis 400 Herzinfarkten pro 100000 Personen der Bevölkerung in Deutschland pro Jahr zeigt die hohe gesundheitspolitische Bedeutung der Vorsorge. Die Betrachtung des Lärms als möglicher Kofaktor bei der Pathogenese des Herzinfarktes beziehungsweise des plötzlichen Herztodes bildete den Schwerpunkt der epidemiologischen Lärmstudie in Berlin. Die vorliegende Arbeit beschäftigte sich dabei vorwiegend mitdem Zusammenhang zwischen der Lärmempfindlichkeit und der Lärmbelästigung durch STrassenverkehrslärm. In den Analysen wurde die Korrelation zwischen beiden Variablen und die mögliche Beeinflussung dieser Korrelation durch soziodemografische Faktoren bestimmt. Im Rahmen einer Fall-Kontroll-Studie wurden für den Zeitraum von zwei Jahren die Angaben von 2235 Probanden ausgewertet. Die Datenerhebung fand im Rahmen eines circa einstündigen Interviews im Krankenhaus statt. Die Gruppe der Fälle bildeten die Patienten mit akuten Herzinfarkt. Die Kontrollgruppe setzte / The high incidence of heart attacks in Germany (approximately 300 to 400 cases for every 100000 people) emphasises the importance of heart attack prevention as a public health issue.Consequently, an epidemiologic study investigating chronic noise as a cofactor of the pathogenesis of the heart attack or sudden heart death is being conducted at the Charite University Hospital, Berlin. This thesis describes the relation between noise annoyance through road traffic noise and noise sensitivity as part of the aforementioned study. The correlations were determined by analysis between both variables while taking into account the possible influence of socialogical factors such as age, sex, lifestyle.To this end, data from 2235 persons were evaluated in a case-control-study within a two year period. Collection and evaluation of the data took place in the hospital and included a 1 hour 1 interview with the subjects. The subjects were taken from a pool of patients with current acute heart attacks, defined as having a acute heart attack within the period of two to ten days previous to the interview. Surgical patients without current heart complications were used as the control study group. Noise annoyance in the daytime and noise sensitivity had a little correlation of rp = 0,23. Correlation was rp = 0,19 between noise annoyance in the nighttime and noise sensitivity. 19 percent of variations of noise annoyance in the daytime were explained by noise sensitivity, sex, income and marital status in a multivariate analysis. 18 percent of variations noise annoyance in the nighttime were explained by noise sensitivity, age and education. 17 percent of variations were explained by noise annoyance and all socialogical factors in the case of dependence of noise sensitivity. There is a significant, but only little correlation between noise sensitivity and noise annoyance. An independent and separate inspection of both factors should be made in clinical studies of heart and circulation diseases, which include noise sensitivity and noise annoyance.

Annoyance due to relatively high levels of sound and noise, above 50 dB, has been well documented in noise assessment literature. The potential for annoyance or disturbance from low amplitude sound, below 50 dB to the threshold of an individual’s hearing, is not as well documented. The thesis presents a new approach to the measurement and assessment of intrusive noise and low amplitude sound. Acoustical and sound quality measures are integrated with measures of loudness, pitch, dissonance and tonality to provide physical measures of sound. Individual amenity is assessed with respect to personal noise sensitivity and personal attitudes to sound in the environment, the environment itself and the perceived qualities of the audible sound. A decision-support methodology to integrate perceived noise with noise performance indicators, annoyance criteria, personal noise sensitivity and amenity is presented. A method for rating intrusive noise is derived. Designs for sound measurement and calibration instrumentation are described. Methods to measure and assess low amplitude sound are presented. Keywords annoyance, intrusive-noise, noise-sensitivity, sound-quality, soundscape

Dans cette thèse, nous étudions trois modèles de percolation planaire : la percolation de Bernoulli, la percolation de Voronoi, et la percolation de lignes nodales. La percolation de Bernoulli est souvent considérée comme le modèle le plus simple à définir admettant une transition de phase. La percolation de Voronoi est quant à elle un modèle de percolation de Bernoulli en environnement aléatoire. La percolation de lignes nodales est un modèle de percolation de lignes de niveaux de champs gaussiens lisses. Deux fils conducteurs principaux ont guidé nos travaux. Le premier est la recherche de similarités entre ces modèles, en ayant à l'esprit que l'on s'attend à ce qu'ils admettent tous la même limite d'échelle. Nous montrons par exemple que le niveau critique de la percolation de lignes nodales est égal au niveau auto-dual (à savoir le niveau zéro) lorsque le champ considéré est le champ de Bargmann-Fock, qui est un champ gaussien analytique naturel. Le deuxième fil conducteur est l'étude de dynamiques sur ces modèles. Nous montrons en particulier que, si on considère un modèle de percolation de Voronoi critique et si on laisse les points se déplacer selon des processus de Lévy stables à très longue portée, alors il existe des temps exceptionnels avec une composante non bornée / We study three models of percolation in the plane: Bernoulli percolation, Voronoi percolation, and nodal lines percolation. Bernoulli percolation is often considered as the simplest model which admits a phase transition. Voronoi percolation is a Bernoulli percolation model in random environment. Nodal lines percolation is a level lines percolation model for smooth planar Gaussian fields. We have followed two main threads. The first one is the resarch of similarities between these models, having in mind that we expect that they admit the same scaling limit. We show for instance that the critical level for nodal lines percolation is the self-dual level (namely the zero level) if the Gaussian field is the Bargmann-Fock field, which is natural analytical field. The second main thread is the study of dynamics on these percolation models. We show in particular that if we sample a critical Voronoi percolation model and if we let each point move according to a long range stable Lévy process, then there exist exceptional times with an unbounded cluster

[pt] Nós começamos essa dissertação com um panorama geral e introdutório dos tópicos de Sensibilidade a Ruído e Percolação . Como essas áreas podem ser desconhecidas por muitos estudantes de pós-graduação, nós apresentamos o material de uma maneira acessível, com o intuito de divulgar importantes técnicas e resultados dessas áreas. Nós também vamos introduzir o modelo para Percolação de Voronoi e apresentar resultados sobre probabilidades de cruzamentos neste modelo. Nos últimos dois capulos nós iremos considerar Sensibilidade a Ruído para Percolação do tipo quenched. Em particular, no penúltimo capítulo nós vamos apresentar resultados do artigo Quenched Voronoi Percolation de Daniel Ahlberg, Simon Griffiths, Robert Morris e Vincent Tassion, e no último capítulo provaremos um teorema que melhora uma das cotas deste artigo. / [en] We begin this dissertation by giving an introductory overview of the topics of Noise Sensitivity and Percolation. As these areas may be unfamiliar to many graduate students, we present the material in an accessible way, with the objective of publicising important techniques and results in these areas.We shall also introduce the model of Voronoi Percolation and present results of Vincent Tassion on crossing probabilities in this model. In the last two chapters we consider Noise Sensitivity of Quenched Voronoi Percolation. In particular, in the penultimate chapter we present the results of the paper Quenched Voronoi Percolation by Daniel Ahlberg, Simon Griffiths, Robert Morris and Vincent Tassion, and in the final chapter we prove a theorem which improves one of the bounds of that paper.

Le bruit de la circulation routière, et en particulier le bruit des deux-roues motorisés, constituent une importante source de gêne sonore. Afin d’estimer l’exposition sonore dans les villes de plus 100 000 habitants, la directive européenne 2002/49/CE impose la réalisation de cartes de bruit stratégiques, basées sur l’indice Lden. Cet indice est également utilisé dans des relations exposition-réponse, afin de prédire les pourcentages de personnes gênées, notamment par le bruit du trafic routier. En couplant les cartes de bruit stratégiques et ces relations exposition-réponse, des cartes de gêne pourraient être établies. Toutefois, la pertinence de cet indice pour prédire la gêne due au bruit en milieu urbain est souvent remise en cause, car de nombreux facteurs acoustiques influents (e.g. les caractéristiques spectrales et temporelles) ne sont pas pris en compte par cet indice. Cette thèse vise à améliorer la caractérisation de la gêne due au bruit de trafic routier urbain en considérant différentes compositions de trafic et la présence des deux-roues motorisés. Dans ce but, des expériences sont menées en conditions contrôlées. Une première étude a porté sur l’influence de plusieurs facteurs acoustiques relatifs aux périodes de calme et aux bruits de passage de véhicules sur la gêne due au bruit de trafic routier urbain. Cette étude a conclu à l’influence de la présence de périodes de calme et du nombre de véhicules au sein du trafic routier urbain et à l’absence d’influence de l’ordre des véhicules routiers, de la position et de la durée des périodes de calme. Ces résultats ont été utilisés afin de mener la caractérisation physique et perceptive de différentes compositions de trafic routier urbain. La régression multi-niveau a été utilisée pour calculer la gêne, en considérant 1) des facteurs acoustiques influents à l’aide de combinaisons pertinentes d’indices et 2) un facteur non acoustique : la sensibilité au bruit. Dans les villes, le bruit routier est souvent entendu en situation de multi-exposition avec d’autres bruits. Dans le cadre de ces travaux de thèse, les situations de multi-exposition aux bruits routier et d’avion ont été étudiées. Pour cela, un travail semblable à celui mené pour le bruit de trafic routier urbain a été mené pour le bruit d’avion conduisant également à des combinaisons pertinentes d’indices. En vue de caractériser les gênes dues aux bruits de trafic routier et d’avion pour des situations de multi-exposition sonore, les données des précédentes expériences ainsi que celles d’une expérience conduite en situation de multi-exposition à ces bruits combinés ont été utilisées au travers d’une régression multi-niveau adaptée, comme cela a pu être mené dans la littérature. La régression multi-niveau a ainsi permis la proposition de modèles de gêne pour chaque source de bruit. Puis, la gêne totale due à des situations de multi-exposition à ces bruits a été étudiée, afin de mettre en évidence les phénomènes perceptifs mis en jeu. Des modèles de gêne totale ont été proposés, en utilisant les modèles de gêne due à chaque source. Enfin, les modèles de gêne obtenus pour chaque source et les modèles de gêne totale ont été confrontés aux données d’une enquête socio-acoustique. A cet effet, une méthodologie a été proposée afin d’estimer les différents indices des modèles à partir des valeurs du Lden, issues de cartes de bruit et utilisées pour définir l’exposition au bruit des personnes enquêtées. Cette confrontation a montré que les modèles proposés à partir d’expériences menées en laboratoire et couplés à la méthodologie d’estimation des indices à partir des valeurs du Lden permettent une bonne prédiction de la gêne in situ. / Road traffic noise, and in particular powered two-wheeler noise, constitute an important source of noise annoyance. In order to estimate the noise exposure in cities of more than 100 000 inhabitants, the European directive 2002/49/EC requires the elaboration of strategic noise maps, based on the Lden index. This index is also used in exposureresponse relationships, to predict the percentages of annoyed people, by road traffic noise for example. By coupling strategic noise maps and these exposure-response relationships, noise annoyance maps could be established. The relevance of this index to predict noise annoyance in cities is however often questioned, since many influential acoustical factors (e.g. spectral and temporal features) are not considered by this index. The aim of this thesis is to enhance the characterization of noise annoyance due to different compositions of urban road traffic including powered two-wheelers. To achieve this goal, experiments were carried out under controlled conditions. A first study concerned the influence of several acoustical features related to quiet periods and vehicle pass-by noises on the annoyance due to urban road traffic noise. This study demonstrated the influence of the presence of quiet periods and of the number of vehicles within the urban road traffic and to the absence of the influence of the order of the vehicle pass-by noises, the position and duration of quiet periods. These results were used to carry out the physical and perceptual characterization of different compositions of urban road traffic noise. Multilevel regression was used to calculate noise annoyance, by coupling combinations of indices relating to influential acoustical features and an individual factor: noise sensitivity. In cities, road traffic noise is often combined with other noises. In the framework of this thesis, noise exposure to road traffic noise combined with aircraft noise was studied. Therefore, the same work as the one performed for urban road traffic noise was carried out for aircraft noise, leading also to relevant combinations of noise indices. In order to characterize annoyances due to road traffic noise and to aircraft noise in a combined exposure situation, data from the previous experiments and from an experiment dealing with these combined noises were used through an appropriate multilevel regression, as done in literature. The regression allows annoyance models for each noise source to be proposed. Then, total annoyance due to combined noises was studied, in order to highlight the perceptual phenomena related to the combined exposure. Total noise annoyance models were proposed, using proposed annoyance model of each noise source. Finally, these single source annoyance models and total annoyance models were tested using data of a socio-acoustic survey. To do this, a methodology has been proposed to estimate the different indices involved in the annoyance models, from the Lden values obtained from the strategic noise maps and used to define the noise exposure of the respondents. This confrontation showed that the models proposed on the basis of experiments carried out under laboratory conditions and coupled with a methodology of estimation of the noise indices from Lden values, enabled a good prediction of in situ annoyance.