This project is a trans-disciplinary effort toward inference and prediction of complex physical systems. Such systems are often ineffectively described by first principles models and should be modeled via a data-driven approach. However, difficulties arise from the high dimensional and multi-scale nature of these systems. Further, only limited and poorly informative observations are typically available. Prototypical of these situations is subsurface ocean inference or the prevention of seizures in neurosciences. For many applications however, some degree of expertise is available. The goal of this project is to leverage both the theoretical and the data science pillars to infer computable models informed from the existing prior knowledge (Physics first principles and theories) and providing new hints into the principles satisfied by the proposed abstractions, amenable to interpretation and refutation. More precisely, building upon the pluridisciplinary expertise of the team in the domains of Fluid Mechanics and Deep Neural Networks (DNN), the goal of the proposal is: - i) to make the model space (neural architecture and computational flow) compliant with the known Physics of the system under consideration, - ii) to exploit the data and inference tools to train efficient models built on first principles, thereby enhancing their robustness and reducing their data-hunger, - iii) to form and inspect the abstractions built by the DNN systems, to check whether they satisfy the expected properties and understand the properties they satisfy. Methods and tools will be first developed with low-dimensional dynamical systems but will then be illustrated and demonstrated on a full-scale turbulent fluid flow numerical simulation.

This project proposes pedagogical innovations and experimentations for the use of gesture in learning intonation control, through computer-human interfaces. Mastering vocal source modulation through laryngeal control – “intonation” in a broad sense, encompassing here fundamental frequency, voice quality, melodic rhythm – is fundamental for vocal communication. Mastering phonatory control is not generally an issue during the acquisition of the native language (L1), since it remains largely unconscious. In contrast, mastering phonatory control is a long, voluntary, and often difficult process to acquire a foreign language (L2), or to control a vocal prosthesis after laryngeal surgery. It has been demonstrated that the use of gestures that mimic speech intonation improves the process of learning intonation control, by coding information through different modalities (auditory, visual, and kinaesthetic). This project aims at taking multimodality integration a step further by exploiting gestural control of voice synthesis. The latter, called chironomy, is a novel research paradigm in human-computer interaction that allows to generate intonation trajectories in real-time, from manual gesture. The produced intonation contour is either transmitted to a voice synthesiser, to control a vocal instrument with gesture; or is injected into an excitation source that is placed into the user’s mouth. This excitation is naturally combined with the user articulation to produce an integrated semi-synthetic voice, this is vocal substitution. Thus, this project aims at investigating the use of such systems for the learning of intonation control in two complementary approaches: 1) Learning the natural control of intonation contours with the help of chironomy for foreign language acquisition; 2) Learning the chironomic control of intonation contours with the help of native language knowledge, for vocal substitution. This project is divided in four scientific tasks. The first relates to the development of performative synthesis tools for intonation learning. The existing chironomic systems use voice synthesis and were proven to be operational for singing, and for controlling Mandarin tones in a pilot study. These systems will be adapted to fit the applicative environments of the project. Moreover, a voice excitation source that is suitable for injection will be developed for vocal substitution. The second task tackles the methodological aspects of the project, and aims at identifying the target intonation patterns of the several languages that will be considered in the project (French, English, and Mandarin), along with their associated gestures. A multimodal corpus will be recorded, and preliminary evaluation of chironomic control will be undertaken. The third task consists in applying these methodological aspects to the evaluation of chironomy in different learning situations with subjects who are learning foreign languages. An experimental protocol will be designed to observe whether chironomy is more effective in a given learning situation as compared to spatial hand gestures or auditory feedback alone. This experiment will be conducted in classroom conditions, on the learning of: lexical tones in Mandarin; intonation of French; and intonation of English, as foreign languages. The fourth task will demonstrate that chironomic control of intonation can be used as a substitution of one’s natural voice. This unique paradigm combines the gestural control of an excitation source and the natural articulation of the user, and the coordination of the two controls will be first investigated. Then a large-scale experiment will be deployed to validate the feasibility of such control in vocal substitution. Overall, this project is the first proof of concept of the use of chironomy in the learning of intonation, which has a potentially high short-term social and cultural impact, both in computer-assisted education and clinical applications.

The sound environment is an important element of quality of life in urban and peri-urban areas. Among the various types of environmental noise, the aeroacoustic noise produced by the interaction of flows and obstacles can be an important source of annoyance. Theses noisy interactions are related to transport noise (landing aircraft, trains, etc.) but also to a more recent problem concerning the noise emitted by the facade elements of modern or renovated buildings. To identify and understand these sources of noise in order to reduce them, mock-ups or samples of obstacles in anechoic wind tunnel are often studied. Sound source imagery is a classic tool for such studies, obtained by using microphone arrays and data processing tools such as beamforming. These techniques are currently developed in very simplified propagation, flow and directivity hypotheses, in order to obtain an analysis that reduces to a two-dimensional vision of the phenomena. Thus, the aim of the MAMIES project is to develop innovative experimental techniques for the study, identification and three-dimensional (3D) imaging of this type of aeroacoustic source in the context of wind tunnel measurements, in configurations for which classical techniques fail: highly 3D configurations, complex and / or unsteady flows, presence of diffracting objects in the noise production zone, complex source directivity. The originality of the project is based on two advances in the identification of environmental aeroacoustic noise: the development of a new generation microphone array (with more than 1,000 microphones) based on MEMS technology, combined with new treatment methods based on the principle of time reversal and associating measurements and numerical simulations. The MAMIES project brings together two partners: the PPRIME Institute (University of Poitiers) where experiments will be carried out in the wind tunnel, and the Jean-Rond d'Alembert Institute (Sorbonne University) where the antenna will be developed. The processing techniques and the production and analysis of the experimental data will be carried out jointly. The developed array will completely encompass the area of the flow containing the noise sources to be identified. Thanks to the large number of microphones the 3D sampling of the radiated acoustic field will give an optimal spatial resolution of the sources. The processing techniques will make it possible to discard the assumptions about the propagation medium because they will be based on a 3D numerical simulation whose input data will be the experimental data, but also the real geometry and flow. This approach will take into account an arbitrary flow, the presence of objects in the flow, and operate in the time domain, allowing the study of transient annoying phenomena from an auditory point of view (eg gust effects). Thus real progress is expected in the ability of analysis of sources of noise. Once the tool is completed and validated, experimental campaigns will be carried out in the wind tunnel, targeting two types of applications. The first concerns the aeroacoustic noise emitted by aircraft wings. A finite wall-mounted airfoil will be studied in depth, then the case of three interacting airfoils which provides a model for high-lifted wings producing a very complex aeroacoustic radiation. The second application concerns the noise of facade elements in modern architectural projects, first through interacting elementary objects, then through the study of real samples.

When we think about great architectural achievements in European history, such as ancient amphitheatres or gothic cathedrals, their importance is strongly tied to their acoustic environment. The acoustics of a heritage site is an intangible consequence of the space’s tangible construction and furnishings. It is ephemeral, while also a concrete result of the physical nature of the environment. Through the “Past Has Ears” project (the PHÉ project), we will explore how via measurements, research, and virtual reconstructions the acoustics of heritage spaces can be documented, reconstructed, and experienced for spaces both existing and in various altered states. Inspired by the project’s namesake (Phé, for the constellation Phoenix), and the relatively recent fires at Notre Dame Cathedral in Paris (2019) and Teatro La Fenice opera hall (1996, also meaning Phoenix), the PHÉ project focuses on the preservation, conservation, and reconstruction of heritage sites, bringing them back from the ashes for use by researchers, stake holders, cultural institutions, and the general public. Comprising research teams with experience in acoustic reconstructions and historical research, paired with national heritage monuments of acoustic importance, the consortium will develop a joint methodology for addressing relevant archaeological acoustics issues across Europe with historians of different disciplines. Specialists in tangible/intangible cultural heritage legal issues ensure the viability and longevity of the methodology guidelines. The consortium will prototype next generation exploration tools for presenting digital acoustic reconstructions to scientists and museum visitors alike. Results will be evaluated with associated test heritage sites, created in partnership with stakeholders and experienced content producers. Presentation methods provide first-person in-situ or off-site explorations, with the ability to experience various historical periods. For deteriorated sites, this approach provides access to situations impossible to experience on-site. Additional uses include participative experiences, employing real-time reconstructions for on-site concerts and other events experienced in the heritage acoustics.

The current craze for architectured materials results from 3 factors: 1. Expected exceptional properties: Studies (experimental, numerical, theoretical) show that in addition to mass gain, the presence of an internal architecture significantly improves certain properties (energy absorption), or even creates others (invisibility cap); 2. Shape optimization: Mesostructure design algorithms have emerged that allow a more automatic exploration of the links between architecture and resulting properties; 3. Additive manufacturing: Manufacturing techniques and their rapid development now make it possible to produce structures with complex internal architectures. However, the expected exceptional properties occur when the scale of mechanical loading is close to that of the mesostructure. Wave propagation and instabilities are situations where the lack of scale separation is necessary for producing non-standard effects. A streamlined approach to the design of such materials involves an intermediate step in which an equivalent effective medium is substituted to the mesostructure of the material. This effective continuum is first optimized, in order to satisfy a given set of specifications, then deshomogenized to reveal the desired mesostructure. However, the classical framework of homogenization assumes infinite scale separation and is therefore ill-suited to continuous modelling of expected phenomena. Taking into account the effects of the mesostructure within a continuous modeling is the scientific lock this project proposes to remove. The developed approach is based on generalized continuum mechanics supplemented by the use of group theory in order to clarify the role of material symmetries on effective behavior. The framework concerns periodic and pseudo-periodic materials and the considered applications are, on the one hand, control of wave propagation (Axis 1) and, on the other hand, prediction and control of instabilities (Axis 2). In both cases, the architecture of the elementary cell is decisive. Its determination from the target effective properties via an inverse problem of architecture is at the core of Axis 3 of the project. These axes are complemented by a transversal axis linked to the development of adapted experimental methods. In more details: 1. Elastodynamics of architectured materials: Dynamics adds distribution of inertia to the optimization problem that traditionally only deals with distribution of stiffness. Depending on the applications, the various networks may or may not be congruent. 2. Controlled instabilities, obtained by a succession of stable post-bifurcated configurations. This requires an optimization of the crystallographic symmetries of the architectured materials. The applications here concern the adjustment of the multifunctional properties of materials by a change in mesostructure due to instabilities generated by a mechanical loading. 3. The definition of an inverse problem of architecture allowing to determine associated mesostructures, for a set of invariants of the given effective material. This axis aims both at "deshomogenizing" the results obtained in Axes 1 and 2 in order to obtain a real architecture, but also at exploring and classifying mesostructures associated with exotic elastic anisotropies (2D and 3D). 4. Development of experimental methods adapted to architectured materials. Experimental homogenization implies a specific control of boundary conditions. Moreover, instabilities will generate large displacements at the edges requiring the development of appropriate experimental means. This axis will be limited to the static behavior of architectured materials.