Search

This thesis addresses the problem of the automatic generation of purely hexahedral meshes for simulation codes when having a mesh carrying volume fraction data as an input, meaning that there can be several materials inside one cell. The proposed approach should create an hexahedral mesh where each cell corresponds to a single material, and where interfaces between materials form smooth surfaces. From a theoretical standpoint, we aim at adapting and extending state-of-the-art techniques and we apply them on examples, some classically issued from CAD models (and imprinted onto a mesh to obtain volume fractions), some procedurally generated cases and others in an intercode capacity where we take the results of a first simulation code to be our inputs. We first define a metric that allows the evaluation of our (or others') results and a method to improve those; we then introduce a discrete material interface reconstruction method inspired from the scientific visualization field and finally we present an algorithmic pipeline, called {sc ELG}, that offers a guarantee on the mesh quality by performing geometrical and topological mesh adaptation.; Ces travaux abordent le problème de la génération automatique de maillages hexaédriques pour des codes de simulation, à partir d'un maillage portant des fractions volumiques, c'est-à-dire dont les mailles peuvent contenir plusieurs matériaux. La solution proposée doit contruire un maillage hexaédrique dans lequel chaque maille correspond à un seul matériau, et dont les interfaces entre matériaux doivent former des surfaces lisses. D'un point de vue théorique, nous cherchons à adapter et étendre des solutions existantes, et à les appliquer sur une large variété d'exemples : certains issus de modèles de CAO (plaqués sur un maillage pour obtenir des fractions volumiques), d'autres générés procéduralement et enfin d'autres utilisés dans un rôle d'intercode, récupérés en sortie de codes de simulation. Nous définissons une métrique permettant d'évaluer notre (et d'autres) méthodes, ainsi qu'un post-processus pour améliorer ces résultats; nous introduisons également une méthode de reconstruction d'interfaces discrètes inspirés de méthodes issues du domaine de la visualisation scientifique, et nous proposons un algorithme appelé {sc ELG} avec garantie sur la qualité du maillage, faisant intervenir des modifications géométriques et topologiques sur ce maillage.

Background The binaural masking level difference (BMLD) has been shown to be constant (10−15dB) for masker spectrum levels from 70dB/Hz down to 30−40dB/Hz and to gradually decrease with lower levels (McFadden, 1968; Hall and Harvey, 1984). The decrease at low levels was larger in an asymmetric condition where the masker was attenuated in only one ear. McFadden predicted the data by assuming that an external and an internal noise would interaurally decorrelate the internal representations of the stimuli. In the present study, the role of nonlinear cochlear processing and asymmetric masker level on the BMLD was investigated using an equalization−cancelation (EC) based binaural model framework. Methods The BMLD was measured for 500−Hz target tones presented in 3−kHz−wide maskers. BMLDs were obtained as a function of masker level in one symmetric and two asymmetric masker conditions: (i) NoSπ: thresholds were measured for masker spectrum levels between 50 and -10dB/Hz, with the same level at both ears; (ii) No′Sπ′50: same as first condition but the masker was attenuated in one ear only and fixed at 50dB/Hz in the non−attenuated ear; (iii) No′SΠ′20: same as second condition but with a masker level of 20dB/Hz in the non−attenuated ear. An EC based binaural model with a frontend including nonlinear peripheral processing (Jepsen et al., 2011) was used to predict these results. Results The BMLD obtained in the No′Sπ′50 condition was smaller than that obtained in the NoSπ condition at all masker levels between 50 and -10dB/Hz. The difference in BMLD between the two conditions gradually increased with decreasing masker level from 50 to 20dB/Hz and remained constant between 20 and -10dB/Hz. The proposed model could account for these data. A model analysis suggested that the increase in BMLD difference between the No′Sπ′50 and NoSπ conditions results from a decrease in interaural correlation at the output of the periphery, and is a consequence of the cochlear nonlinearity at levels between 20 and 50dB/Hz. For levels below 20dB/Hz, cochlear processing becomes linear and the model predicts that the difference in BMLD between the symmetric and the two asymmetric conditions is constant, in line with the experimental data. Conclusion A model was proposed to account for the effect of level asymmetry on BMLD. The modeling results suggest that cochlear nonlinearites affect the analysis of binaural cues a at higher processing stages such that signals carrying interaural level differences can become interaurally decorrelated.

Geophysical Journal International, 186 (2), ISSN:0956-540X, ISSN:1365-246X

Humans are highly sensitive to Interaural Time Differences (ITDs) in stimuli presented via headphones. For broadband noise stimuli of long durations, ITD detection thresholds can be as low as 10 to 15 ?s. When the stimulus duration is shortened, thresholds increase by about a factor 2 for a tenfold decrease in duration. ITD thresholds also increase, when the probe carrying an ITD is surrounded by diotic fringes. When a 5-ms probe is combined with preceding or trailing fringes, the effect of a fringe preceding the probe is stronger than that of a trailing fringe for fringe durations < 35 ms. The effect of fringes surrounding the probe is equal to the addition of the effects of the individual fringes. In this contribution, we present behavioral data for the same experimental condition, called dynamically varying ITD detection, but for a wider range of probe and fringe durations. Probe durations varied between 5 and 400 ms, and fringe durations had values of 5, 20, 100 or 200 ms. In contrast to earlier findings, we observed for most duration combinations a stronger effect of the trailing fringe than of the preceding fringe. For these configurations, the effect of surrounding fringes was dominated by the trailing fringe. Only for the combination of 5-ms fringes with 5-ms probes did we see the clear dominance of the preceding fringe. These results are not easy to align with the concept of onset emphasis often used to explain binaural localization data for short stimuli. In fact the data seem to be difficult to predict with a purely signal-driven model of perception and thus form an interesting challenge for modeling human localization.

Humans are highly sensitive to Interaural Time Differences (ITDs) in stimuli presented via headphones. For broadband noise stimuli of long durations, ITD detection thresholds can be as low as 10 to 15 ms. When the stimulus duration is shortened, thresholds increase by about a factor 2 for a tenfold decrease in duration. ITD thresholds also increase, when the probe carrying an ITD is surrounded by diotic fringes. When a 5-ms probe is combined with preceding or trailing fringes, the effect of a fringe preceding the probe is stronger than that of a trailing fringe for fringe durations

International audience