Search

International audience; Natural and semi-natural grasslands, particularly dry grasslands, are high-value biodiversity hotspots in Europe and major challenges for ecological conservation and restoration. Soil transfer has been used successfully to restore dry grassland ecosystems (soil and vegetation), but should only be applied as a last resort and where an undegraded ecosystem has previously independently been earmarked for destruction. Assessing overall restoration success means taking soil recovery into account as well as vegetation regeneration. In Mediterranean regions and under global warming, soil hydrodynamic properties are the most critical limiting factors for the restoration of dry ecosystems. This study analyses soil physico-chemical parameters and hydrodynamic properties in reconstructed soils after soil transfer following a petrol land spill. Six years after soil transfer and compared to the surrounding soil reference steppe, the horizons of the soil transfer treatments were significantly less colonised by plant roots and showed very few biological activities and earthworm galleries. Then, their porosity was also significantly lower. Hydrodynamic properties continue to differ significantly between the restored site and the surrounding reference steppe ecosystem. There is also less available soil water and water drains faster in all soil transfer treatments. This can be explained by significant higher finer grain size (clays, silt), chemical elements (N, P, K), and plant cover for soil transfer treatments but significant, lower levels of organic matterdue to mechanical soil compaction during soil spreading. The soil transfer treatment involving full reconstitution of the three main soil layers yields the most encouraging results. However, the question arises of whether long-term vegetation recovery might be compromised by soil hydrological dysfunction under global warming.

International audience; Extensive green roofs have been shown to support native biodiversity and plant communities that are analogous to natural or semi-natural habitats such as grasslands. However, little is known about the role of soil seedbanks in the dynamic of extensive green roof plant communities. The purpose of this study was to analyze the seedbank that developed after 4 years of an extensive green roof analog to dry grassland plant community, seeded with 29 species. We aimed to understand the contribution of seedbank to the resilience of vegetation to harsh conditions of the roof and to colonization by surrounding spontaneous species. We monitored the plant species cover in 36 plots during 4 years in June (between 2018 and 2021), and sampled the seedbank in February 2021. Our results showed that the soil seedbank was dominated by transient spontaneous ruderals species, while the standing vegetation was still dominated by seeded grassland species. We found that seeded grassland species had poor seedbank stock, similar to their natural environments. The similarity index between the standing vegetation and the seedbank increased over time, and we measured a significant correlation between dominant species cover and their seedbank density. Spontaneous species cover was not correlated to the proportion of soil not colonized by seeded species cover, indicating that gaps in vegetation did not influence the development of spontaneous species. Our findings highlight the importance of seedbank in the dynamic of green roof vegetation and demonstrate that analogous habitat species exhibit similar behavior as in their natural environments.

IntroductionSemi-natural dry grasslands are among the species-richest and most diverse habitats in Europe especially in the Mediterranean Basin. They evolved with long-term severe environmental constraints, either biotic (e.g., grazing or mowing) or abiotic (e.g., dryness or oligotrophy). Their recovery after sever anthropogenic disturbance, such as cultivation is often very slow and incomplete. A deeper understanding of the processes driving the recovery would be of great benefit to better restore or conserve these grasslands. We used the filter model as a framework to assess the relative importance of dispersion, abiotic and biotic filters as drivers of this recovery.MethodsWe studied the medium-term dynamics of 37 Mediterranean dry grassland plant communities after an episode of cultivation in the plain of La Crau (Southern France) by two census: after 25 years of abandonment, and revisited 10 years later (35 years of abandonment). Our aims were to characterize the temporal medium-term dynamics of these communities, to measure their taxonomic resilience and to identify the drivers. We performed structural equation modeling to measure the importance of the various drivers, with cover of remnant dry grasslands surrounding each studied community as a proxy for the dispersion filter, soil pH and phosphorus content as proxies for the abiotic filter and cover of the most dominant species as a proxy for the biotic filter.Results and DiscussionOur results show that former arable field vegetation do not recover after more than 40 years. The three filters drive the recovery of dry grassland vegetation: former arable fields that have lower phosphorus content (abiotic filter) have a lower cover of dominant species (biotic filter), a higher species richness and a higher similarity with remnant dry grassland plant communities, especially if the former arable fields are surrounded by remnant dry grasslands (dispersion filter).

International audience; La steppe de Crau est un des joyaux naturels de la Provence que le Conservatoire d'espaces naturels de Provence-Alpes-Côte d'Azur contribue à protéger depuis plus de 40 ans. Sous son apparence aride, monotone et inhospitalière, le coussoul constitue un écosystème unique que les scientifiques s'emploient aussi à réparer. Cet article retrace 20 années de recherches en « écologie de la restauration » effectuées par le CNRS et ses partenaires sur le territoire de la plaine de Crau.

International audience; Tout le monde s’accorde à dire qu’une planète en bonne santé est une condition nécessaire au bien-être de l’humanité. Mais quels sont les moyens d’action mis à disposition de nos sociétés pour s’engager dans cette nécessaire transition écologique ? Comment changer nos comportements et agir au mieux pour le vivant… et donc aussi pour nous ? Quatre chercheurs, tous écologues, portent ici leur regard sur les relations qu’entretiennent l’homme, la société et l’environnement. Éclairants, enrichissants et facilement accessibles, leurs propos nous offrent ainsi de nouvelles perspectives et nous invitent à repenser nos pratiques tout en nous offrant un beau voyage à la découverte de leurs univers de recherche.

Sketch-Based Image Retrieval (SBIR) is a crucial task in multimedia retrieval, where the goal is to retrieve a set of images that match a given sketch query. Researchers have already proposed several well-performing solutions for this task, but most focus on enhancing embedding through different approaches such as triplet loss, quadruplet loss, adding data augmentation, and using edge extraction. In this work, we tackle the problem from various angles. We start by examining the training data quality and show some of its limitations. Then, we introduce a Relative Triplet Loss (RTL), an adapted triplet loss to overcome those limitations through loss weighting based on anchors similarity. Through a series of experiments, we demonstrate that replacing a triplet loss with RTL outperforms previous state-of-the-art without the need for any data augmentation. In addition, we demonstrate why batch normalization is more suited for SBIR embeddings than l2-normalization and show that it improves significantly the performance of our models. We further investigate the capacity of models required for the photo and sketch domains and demonstrate that the photo encoder requires a higher capacity than the sketch encoder, which validates the hypothesis formulated in [34]. Then, we propose a straightforward approach to train small models, such as ShuffleNetv2 [22] efficiently with a marginal loss of accuracy through knowledge distillation. The same approach used with larger models enabled us to outperform previous state-of-the-art results and achieve a recall of 62.38% at k = 1 on The Sketchy Database [30].

Stereo vision is essential for many applications. Currently, the synchronization of the streams coming from two cameras is done using mostly hardware. A software-based synchronization method would reduce the cost, weight and size of the entire system and allow for more flexibility when building such systems. With this goal in mind, we present here a comparison of different deep learning-based systems and prove that some are efficient and generalizable enough for such a task. This study paves the way to a production ready software-based video synchronization system., Comment: 5 pages, 4 figures, Accepted at ICASSP 2023

Cardiotocography (CTG) is a key element when it comes to monitoring fetal well-being. Obstetricians use it to observe the fetal heart rate (FHR) and the uterine contraction (UC). The goal is to determine how the fetus reacts to the contraction and whether it is receiving adequate oxygen. If a problem occurs, the physician can then respond with an intervention. Unfortunately, the interpretation of CTGs is highly subjective and there is a low inter- and intra-observer agreement rate among practitioners. This can lead to unnecessary medical intervention that represents a risk for both the mother and the fetus. Recently, computer-assisted diagnosis techniques, especially based on artificial intelligence models (mostly supervised), have been proposed in the literature. But, many of these models lack generalization to unseen/test data samples due to overfitting. Moreover, the unsupervised models were applied to a very small portion of the CTG samples where the normal and abnormal classes are highly separable. In this work, deep unsupervised learning approaches, trained in a semi-supervised manner, are proposed for anomaly detection in CTG signals. The GANomaly framework, modified to capture the underlying distribution of data samples, is used as our main model and is applied to the CTU-UHB dataset. Unlike the recent studies, all the CTG data samples, without any specific preferences, are used in our work. The experimental results show that our modified GANomaly model outperforms state-of-the-arts. This study admit the superiority of the deep unsupervised models over the supervised ones in CTG abnormality detection.

For each restoration project, it is essential to define a reference ecosystem towards which ecological restoration aims. Historical information may be useful in the construction of the reference model and in finding out about the type and degree of disturbances that have caused degradation on the site, as legacies of past human land-uses have been shown to be major determinants of vegetation and plant diversity even in the very long term. The aim of this paper is the retrace the history of three sites where restoration protocols were tested, to better apprehend the potential reference ecosystems as well as the disturbances that degraded them. To do so, we first compiled information on the reference ecosystems at the landscape scale. Then, we compiled historical and agricultural information at the scale of the sites from local historical documents, historical reviews and aerial photographs. The three sites most likely were grasslands or oak savannas before European settlement. Grasslands used to cover 8.9 million ha in California. They were species-rich grasslands managed with fire by the native people and grazed by native ungulates. Early settlers introduced cattle grazing and changed fire regimes. Later on, many grasslands were converted to cultivation. All three sites were used for ranching and sometimes overgrazed. At least two of them were cultivated, probably leading to restoration being harder there, as plowing eliminates perennial species, destroys the seedbank, changes soil characteristics., Buisson Elise,Dutoit Thierry. Restoration sites on the Californian coast have different ecological histories that can influence restoration success. In: Ecologia mediterranea, tome 48 n°1, 2022. pp. 33-45.

International audience; This paper presents a new guitar dataset made out of richly annotated guitarist improvisations. The annotations include notes, playing techniques, instrument tuning, audio effects configurations as well as transcriptions of post improvisation interviews. The dataset gathers ten hours of improvisations and around five hours of interviews. These accompanying data make this dataset suitable for a variety of research domains : from MIR to musical improvisation analysis and musicology. The recordings that have yielded to this dataset were done in the context of an hexaphonic multieffect practice study. This hexaphonic multieffect is meant to work with an hexaphonic guitar (one pickup per string guitar) and grants the player with independent audio effects configurations for each string. This paper presents the dataset, and the experiment it has been gathered from. It also details, based on the transcriptions of the interviews, a first analysis of the specificities of an hexaphonic multieffect practice.

Organisé dans le cadre du Réseau Thématique Pluridisciplinaire « Paysage et Environnement » du CNRS et dans le but affiché de favoriser les échanges disciplinaires ainsi que la diffusion des recherches de jeunes chercheurs, le colloque « Paysages et environnement. De la reconstitution du passé aux modèles prospectifs » a rassemblé une large communauté scientifique ainsi que de nombreux représentants d’organismes publics impliqués dans l’étude, la gestion ou la conservation des paysages et de l’environnement. Articulé autour de trois thèmes principaux : « Le passé : reconstitutions, transmission et nouveaux indicateurs des dynamiques environnementales » ; « Représentations sociales et formes de l’action » et enfin « Modélisation, scénarios prospectifs et rétro-prospectifs », ce volume réunit trente-six articles. Il brosse un tableau représentatif du dynamisme et de la richesse des recherches menées aujourd’hui dans ces domaines, qu’il s’agisse d’études mono-disciplinaires ou d’approches interdisciplinaires intégrées. Les contributions, originales pour la plupart, résultent de travaux ou de programmes de recherches collectifs faisant appel à des approches traditionnelles ainsi qu’à des démarches innovantes visant à la reconstitution des environnements et des paysages du passé mais aussi, dans le cadre d’approches prospectives, à la détermination de leurs évolutions futures.

Les habitats ouverts du Nord-Ouest de l’Europe sont considérés comme étant le résultat de défrichements au Néolithique ou de l’impact du pâturage d’herbivores sauvages au Mésolithique. Nous avons réalisé une étude dans une pelouse calcaire de la vallée de Seine (Haute-Normandie, France) grâce à l’utilisation de la pédoanthracologie. Elle montre que les essences arborées de fin de succession étaient encore présentes au Néolithique. Open habitats in north-western Europe are considered to be the result of land clearing in the Neolithic or the grazing impact of wild herbivores in the Mesolithic. We carried out a study in a chalk grassland of the Seine valley (Upper-Normandy, France) by using soil wood charcoal analysis. We show that late-successional woody species were still present in the Neolithic.

L’origine des espaces ouverts en Europe fait encore l’objet de nombreuses controverses. Dans un site atelier particulier, la pseudo-steppe de la plaine de Crau dans les Bouches du Rhône, l’absence de colonisation des essences forestières après abandon du pâturage pourrait-elle s’expliquer par les conditions biotiques, climatiques ou édaphiques ? L’identification et la datation de charbons de bois échantillonnés sous les fondations de bergeries romaines et dans les sols steppiques apportent les premiers éléments de réponse. The origine of European open landscapes is still debated. In the pseudo-steppic area of la Crau in south-eastern France, the absence of tree colonisation after sheep grazing abandonment could be explain by biotic, climatic or edaphic conditions ? Identification and datations of wood charcoals sampled under Roman sheepfold foundations bring the first elements of response.

Evaluating the Quality of a Synthesized Motion with the Fr\'echet Motion Distance, Comment: 2 pages, 2 figures

For many years now, understanding the brain mechanism has been a great research subject in many different fields. Brain signal processing and especially electroencephalogram (EEG) has recently known a growing interest both in academia and industry. One of the main examples is the increasing number of Brain-Computer Interfaces (BCI) aiming to link brains and computers. In this paper, we present a novel framework allowing us to retrieve the attention state, i.e degree of attention given to a specific task, from EEG signals. While previous methods often consider the spatial relationship in EEG through electrodes and process them in recurrent or convolutional based architecture, we propose here to also exploit the spatial and temporal information with a transformer-based network that has already shown its supremacy in many machine-learning (ML) related studies, e.g. machine translation. In addition to this novel architecture, an extensive study on the feature extraction methods, frequential bands and temporal windows length has also been carried out. The proposed network has been trained and validated on two public datasets and achieves higher results compared to state-of-the-art models. As well as proposing better results, the framework could be used in real applications, e.g. Attention Deficit Hyperactivity Disorder (ADHD) symptoms or vigilance during a driving assessment.

International audience

National audience

International audience; Bioindication of ecological variables such as humidity, temperature or pH by ecological indicatorvalues of plants is a powerful tool for research in plant ecology, e.g. to detect early vegetationchanges. Here, we provide a data set of ecological indicator values including niche width for anentire regional flora. We used an extensive data-base with floristic relevés from Southern Franceto recalibrate indicator values for light (L), temperature (T), continentality (K), air humidity (A),soil moisture (F), pH (R), productivity (N), soil texture (G), soil organic matter content (O) andsalinity (S). Values were recalibrated using average values from co-occurring plants, enabling todevelop indicator values for species not yet evaluated previously. Recalibrated values are on acontinuous scale and we add standard deviation, median, first and third quartile for each indicatorvalue. Linear regression of average indicator values against measured factors showed highercorrelation with recalibrated values compared to original indicator values for temperature, pH andnitrogen, and comparable R 2 for moisture. Individual indicator systems performed better than acombination and applying different weighting procedures demonstrated the usefulness of inversevariance. We further illustrate graphically how recalibrated values and niche width increaseecological knowledge on plants.

National audience

In the past few years, neural character animation has emerged and offered an automatic method for animating virtual characters. Their motion is synthesized by a neural network. Controlling this movement in real time with a user-defined control signal is also an important task in video games for example. Solutions based on fully-connected layers (MLPs) and Mixture-of-Experts (MoE) have given impressive results in generating and controlling various movements with close-range interactions between the environment and the virtual character. However, a major shortcoming of fully-connected layers is their computational and memory cost which may lead to sub-optimized solution. In this work, we apply pruning algorithms to compress an MLP- MoE neural network in the context of interactive character animation, which reduces its number of parameters and accelerates its computation time with a trade-off between this acceleration and the synthesized motion quality. This work demonstrates that, with the same number of experts and parameters, the pruned model produces less motion artifacts than the dense model and the learned high-level motion features are similar for both, Comment: 8 pages, 4 figures, 2 tables, 17th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications - Volume 1: GRAPP, ISBN 978-989-758-555-5, ISSN 2184-4321, pages 286-293

National audience

International audience; Quarry restoration is difficult to implement due to the scarcity of the original soil. The restoration of Mediterranean dry grasslands to mitigate similar ecosystem/habitat loss is thus still a developing process. So far, very few studies have created artificial substrates to address the lack of soil. In the La Crau plain (Southeastern France), quarries were established in an ancient Mediterranean steppe-like dry grassland. Once the remaining grassland zones were protected in 2001, quarries were extended over former intensive orchards established on the grassland in the 1990s before it was protected. Now, restoration has to be done without the unaltered protected grassland soil but with the soil from orchards, which contains fertilizers spread during the orchard exploitation. To recreate a more suitable substrate, the orchard topsoil was mixed with poor substrate materials directly extracted from the quarry (0–30 mmpebbles with sandy matrix). Different substrate mixtures were tested with or without sowing the dominant species of the grassland, a perennial grass: Brachypodium retusum (and an annual grass B. hybridum). After 5 years of monitoring, our results show that raw substrate materials (100%) limit the establishment of all species, and that, at the opposite, orchard topsoil (100%) favors a significantly higher species richness of target and non-target species. The recommended substrate mixture to reach a compromise between high target species cover and low non-target species cover is composed of 50% raw quarrymaterial and 50% arable soil. Sowing Brachypodium not only limits non-target species richness and cover but also reduces target species richness establishment.

Sketch-based image retrieval (SBIR) is the task of retrieving natural images (photos) that match the semantics and the spatial configuration of hand-drawn sketch queries. The universality of sketches extends the scope of possible applications and increases the demand for efficient SBIR solutions. In this paper, we study classic triplet-based SBIR solutions and show that a persistent invariance to horizontal flip (even after model finetuning) is harming performance. To overcome this limitation, we propose several approaches and evaluate in depth each of them to check their effectiveness. Our main contributions are twofold: We propose and evaluate several intuitive modifications to build SBIR solutions with better flip equivariance. We show that vision transformers are more suited for the SBIR task, and that they outperform CNNs with a large margin. We carried out numerous experiments and introduce the first models to outperform human performance on a large-scale SBIR benchmark (Sketchy). Our best model achieves a recall of 62.25% (at k = 1) on the sketchy benchmark compared to previous state-of-the-art methods 46.2%.

International audience; Ramose false brome, Brachypodium retusum (Pers.) P. Beauv., is a perennial herbaceous species that dominates French dry Mediterranean grasslands. In the Crau plain (South-East of France), its spon-taneous re-establishment is very low after human interventions involving soil disturbance contrary to studies from other regions. The major objective of this PhD thesis was to test different hypothesis potentially explaining the different colonisation patterns for improving the re-establishment of this species in future ecological restoration projects. We tested whether these different colonisation patterns are the result of genetic differentiation among populations. We further studied the adaptive differentiation of phenotypic traits to several environmental factors. The second part of the study analysed the environmental factors that limit in situ re-establishment. We tested the effect of grazing and fire on vegetative cover as well as on sexual reproduction of established B. retusum. Finally, we measured the effect of initial watering and grazing on seedling establishment. Our results showed that populations of B. retusum are genetically differentiated in neutral markers but also in phenotypic traits. This differentiation is superior to drift alone and suggests adaptation to environmental condi-tions. In situ experiments showed that fire has a positive effect on B. retusum reproduction in adult populations and on the associated plant community. During the second season, the positive effect of watering on seedling survival was significant only in the grazed plots. Grazing in early life cycle stages had a negative effect on seedling recruitment and growth.; Le Brachypode rameux, Brachypodium retusum (Pers.) P. Beauv., est une espèce herbacée pérenne qui domine les pelouses sèches méditerranéennes. Dans la plaine de la Crau (Sud-Est de la France), sa recolonisation spontanée est très faible après perturbation du sol. L’objectif principal de ce travail a donc été de tester les différentes hypothèses pouvant expliquer les raisons de cette faible colonisation dans le but d’améliorer l’efficacité des opérations de restauration écologique futures. Nous avons d’abord testé si sa dynamique résulte d’une différenciation génétique entre les populations. Nous avons ensuite étudié la différenciation adaptative des traits phénotypiques en fonction de plusieurs facteurs environnementaux clés et analysé quels facteurs environnementaux limitent alors la recolonisation. Nous avons aussi testé in situ l’effet du pâturage et du feu sur le recouvrement végétatif ainsi que sur la reproduction sexuée de B. retusum. Pour finir, nous avons mesuré l’effet de l’arrosage initial et du pâturage sur l’installation de ses plantules. Nos résultats ont montré que les populations de B. retusum sont génétiquement différenciées pour les marqueurs neutres et les traits phénotypiques. Cette différenciation est supérieure à la dérive seule et suggère donc une adaptation aux conditions environnementales. Les expériences in situ ont, quant à elles, montré que le feu a un effet positif sur la reproduction de B. retusum et sur sa communauté végétale associée. Enfin, l’effet de l’arrosage a été positif sur la survie de ses plantules uniquement dans les parcelles pâturées. Le pâturage au début du cycle de vie a cependant eu un effet négatif sur le recrutement et la croissance des plantules.

Le 07 août 2009, la fuite accidentelle d’un oléoduc provoquait une marée noire terrestre sur plus de 5 hectares dans la plaine de la Crau, au cœur d’une réserve naturelle située dans le sud-est de la France. Devant l’ampleur de la pollution générée par les 4 700 m3 de pétrole brut épandu, 72 000 tonnes de sols pollués sont retirées l’année suivante sur une profondeur moyenne de 50 cm pour être ensuite stockées dans une décharge spécialisée. Depuis, plusieurs méthodes ont été testées pour restaurer les écosystèmes dans la zone. L’une d’entre elles est fondée sur l’introduction de fourmis moissonneuses dans les sols contaminés.

International audience; Les transformations écologiques sont l’empreinte de nos cultures. En tant qu’artiste, me prononcer sur ces changements aurait peut-être l’intérêt d’un erratum dans la rubrique horoscope d’une feuille de chou… Ma démarche suit le parfum des rencontres et s’imprègne des sciences. L’inspiration et l’expression émanent d’interdépendances et interagissent. Ainsi, comment rendre nos pas aussi subtils qu’un brachypode et nous donner le goût de nouer notre liberté dans un maillage du vivant qui s’étend ?

International audience

International audience; 2008-2019, more than a decade of exchanges and debates around ecological restoration in France.The REVER 10 conference “Restoring or recovering” organized by MNHN (National Museum of Natural History) and REVER association in March 2019, was an opportunity to review the development of restoration ecology and ecological restoration, as the “Decade of Ecosystem Res-toration” was being proclaimed by the UN. In France, the development of ecological restoration has been guided by the REVER association, a network created in 2008, on the model of the SER (Society for Ecological Restoration). It contributes to organizing annual conferences, promotes exchanges between technical and scientific stakeholders and the diffusion of knowledge, con-cepts and practices. These exchanges were particularly important at a time when, over the same period, research and feedbacks from restoration projects had been leading to an evolution and a strengthening of concepts through numerous scientific debates around, in particular, the notion of reference ecosystems, novel ecosystems and passive restoration. Today more than ever, even if conservation must remain a priority, the challenges we face in ecological restoration make it es-sential to strengthen both the acquisition of knowledge and restoration actions. Thus, scientific dissemination, the design of evaluation tools and of decision support systems and exchanges around restoration project feedbacks remain essential in which REVER along with international structures have an undeniable role to play.; Le colloque REVER 10 « Restaurer ou reconquérir » organisé en mars 2019 par le MNHN et l’association REVER, a été l’occasion de faire un point d’étape sur le développement de l’écologie de la restauration et de la restauration écologique, au moment où débute la « décennie de la restauration des écosystèmes » décrétée par l’ONU. En France, le développement de la restauration écologique a été accompagnée par le réseau REVER créé en 2008 sur le modèle de la SER (Society for Ecological Restoration) qui favorise, notamment par l’organisation de colloques annuels, les échanges entre acteurs techniques et scientifiques ainsi que la circulation des connaissances, des concepts et des pratiques. Ces échanges ont été particulièrement importants au moment où, sur la même période, la recherche et les retours d’expériences ont conduit à une évolution et à un confortement des concepts par de nombreux débats scientifiques autour notamment des notions d’écosystèmes de référence, de nouveaux écosystèmes et de restauration passive. Aujourd’hui plus que jamais, même si la préservation doit rester une priorité, les enjeux de la restauration rendent indispensable le renforcement tant de l’acquisition de connaissances que de l’action. Ainsi la diffusion scientifique, la construction d’outils d’évaluation et d’aide à la décision et les échanges autour de retours d’expériences restent primordiaux et le réseau REVER, en lien avec les structures internationales, a un rôle indéniable à jouer.

In this paper, we present preliminary results of our ongoing work on cross-corpora analyses of smiles and laughter mimicry. For this, instead of recording new data, we leverage the ones produced and available. We analyze smiles and laughs mimicry in three different datasets and show results similar to our previous work. The data used here can be accessed at: https://doi.org/10.5281/zenodo.3820510., Laughter and Other Non-Verbal Vocalisations Workshop: Proceedings (2020), Laughter and Other Non-​Verbal Vocalisations Workshop 2020

ICE-Talk is an open source web-based GUI that allows the use of a TTS system with controllable parameters via a text field and a clickable 2D plot. It enables the study of latent spaces for controllable TTS. Moreover it is implemented as a module that can be used as part of a Human-Agent interaction.

The comparative study presented in this paper focuses on two approaches for the search of sound presets using a specific geometric touch app. The first approach is based on independent sliders on screen and is called analytic. The second is based on interpolation between presets represented by polygons on screen and is called holistic. Participants had to listen to, memorize, and search for sound presets characterized by four parameters. Ten different configurations of sound synthesis and processing were presented to each participant, once for each approach. The performance scores of 28 participants (not including early testers) were computed using two measured values: the search duration, and the parametric distance between the reference and answered presets. Compared to the analytic sliders-based interface, the holistic interpolation-based interface demonstrated a significant performance improvement for 60% of sound synthesizers. The other 40% led to equivalent results for the analytic and holistic interfaces. Using sliders, expert users performed nearly as well as they did with interpolation. Beginners and intermediate users struggled more with sliders, while the interpolation allowed them to get quite close to experts' results.

This paper addresses the problem of estimating the voice source directly from speech waveforms. A novel principle based on Anticausality Dominated Regions (ACDR) is used to estimate the glottal open phase. This technique is compared to two other state-of-the-art well-known methods, namely the Zeros of the Z-Transform (ZZT) and the Iterative Adaptive Inverse Filtering (IAIF) algorithms. Decomposition quality is assessed on synthetic signals through two objective measures: the spectral distortion and a glottal formant determination rate. Technique robustness is tested by analyzing the influence of noise and Glottal Closure Instant (GCI) location errors. Besides impacts of the fundamental frequency and the first formant on the performance are evaluated. Our proposed approach shows significant improvement in robustness, which could be of a great interest when decomposing real speech.

An inversion of the speech polarity may have a dramatic detrimental effect on the performance of various techniques of speech processing. An automatic method for determining the speech polarity (which is dependent upon the recording setup) is thus required as a preliminary step for ensuring the well-behaviour of such techniques. This paper proposes a new approach of polarity detection relying on oscillating statistical moments. These moments have the property to oscillate at the local fundamental frequency and to exhibit a phase shift which depends on the speech polarity. This dependency stems from the introduction of non-linearity or higher-order statistics in the moment calculation. The resulting method is shown on 10 speech corpora to provide a substantial improvement compared to state-of-the-art techniques.

It was recently shown that complex cepstrum can be effectively used for glottal flow estimation by separating the causal and anticausal components of speech. In order to guarantee a correct estimation, some constraints on the window have been derived. Among these, the window has to be synchronized on a Glottal Closure Instant. This paper proposes an extension of the complex cepstrum-based decomposition by incorporating a chirp analysis. The resulting method is shown to give a reliable estimation of the glottal flow wherever the window is located. This technique is then suited for its integration in usual speech processing systems, which generally operate in an asynchronous way. Besides its potential for automatic voice quality analysis is highlighted.

This paper investigates the differences occuring in the excitation for different voice qualities. Its goal is two-fold. First a large corpus containing three voice qualities (modal, soft and loud) uttered by the same speaker is analyzed and significant differences in characteristics extracted from the excitation are observed. Secondly rules of modification derived from the analysis are used to build a voice quality transformation system applied as a post-process to HMM-based speech synthesis. The system is shown to effectively achieve the transformations while maintaining the delivered quality.

Speech generated by parametric synthesizers generally suffers from a typical buzziness, similar to what was encountered in old LPC-like vocoders. In order to alleviate this problem, a more suited modeling of the excitation should be adopted. For this, we hereby propose an adaptation of the Deterministic plus Stochastic Model (DSM) for the residual. In this model, the excitation is divided into two distinct spectral bands delimited by the maximum voiced frequency. The deterministic part concerns the low-frequency contents and consists of a decomposition of pitch-synchronous residual frames on an orthonormal basis obtained by Principal Component Analysis. The stochastic component is a high-pass filtered noise whose time structure is modulated by an energy-envelope, similarly to what is done in the Harmonic plus Noise Model (HNM). The proposed residual model is integrated within a HMM-based speech synthesizer and is compared to the traditional excitation through a subjective test. Results show a significative improvement for both male and female voices. In addition the proposed model requires few computational load and memory, which is essential for its integration in commercial applications.

This paper addresses the problem of pitch modification, as an important module for an efficient voice transformation system. The Deterministic plus Stochastic Model of the residual signal we proposed in a previous work is compared to TDPSOLA, HNM and STRAIGHT. The four methods are compared through an important subjective test. The influence of the speaker gender and of the pitch modification ratio is analyzed. Despite its higher compression level, the DSM technique is shown to give similar or better results than other methods, especially for male speakers and important ratios of modification. The DSM turns out to be only outperformed by STRAIGHT for female voices.

In the framework of assessing the pathology severity in chronic cough diseases, medical literature underlines the lack of tools for allowing the automatic, objective and reliable detection of cough events. This paper describes a system based on two microphones which we developed for this purpose. The proposed approach relies on a large variety of audio descriptors, an efficient algorithm of feature selection based on their mutual information and the use of artificial neural networks. First, the possible use of a contact microphone (placed on the patient's thorax or trachea) in complement to the audio signal is investigated. This study underlines that this contact microphone suffers from reliability issues, and conveys little new relevant information compared to the audio modality. Secondly, the proposed audio-only approach is compared to a commercially available system using four sensors on a database with different sound categories often misdetected as coughs, and produced in various conditions. With average sensitivity and specificity of 94.7% and 95% respectively, the proposed method achieves better cough detection performance than the commercial system., Comment: arXiv admin note: substantial text overlap with arXiv:2001.00537

This paper focuses on the analysis and synthesis of hypo and hyperarticulated speech in the framework of HMM-based speech synthesis. First of all, a new French database matching our needs was created, which contains three identical sets, pronounced with three different degrees of articulation: neutral, hypo and hyperarticulated speech. On that basis, acoustic and phonetic analyses were performed. It is shown that the degrees of articulation significantly influence, on one hand, both vocal tract and glottal characteristics, and on the other hand, speech rate, phone durations, phone variations and the presence of glottal stops. Finally, neutral, hypo and hyperarticulated speech are synthesized using HMM-based speech synthesis and both objective and subjective tests aiming at assessing the generated speech quality are performed. These tests show that synthesized hypoarticulated speech seems to be less naturally rendered than neutral and hyperarticulated speech.

National audience

As part of the Human-Computer Interaction field, Expressive speech synthesis is a very rich domain as it requires knowledge in areas such as machine learning, signal processing, sociology, psychology. In this Chapter, we will focus mostly on the technical side. From the recording of expressive speech to its modeling, the reader will have an overview of the main paradigms used in this field, through some of the most prominent systems and methods. We explain how speech can be represented and encoded with audio features. We present a history of the main methods of Text-to-Speech synthesis: concatenative, parametric and statistical parametric speech synthesis. Finally, we focus on the last one, with the last techniques modeling Text-to-Speech synthesis as a sequence-to-sequence problem. This enables the use of Deep Learning blocks such as Convolutional and Recurrent Neural Networks as well as Attention Mechanism. The last part of the Chapter intends to assemble the different aspects of the theory and summarize the concepts., 19 pages, 6 figures. To be published in the book "Human Computer Interaction" edited by Prof. Yves Rybarczyk, published by IntechOpen

International audience; Ecological engineering has been originally defined as the design of sustainable ecosystems that integrates human society with its natural environment for the benefit of both (Mitsch and Jørgensen, 1989). This nature-based management might be relevant especially to restore sustainable SUITMA. Despite promising results using earthworms (Ampoorter et al 2011), termites (Mando 1997) or ants (Bulot et al 2014), soil invertebrates that play a major role in ecosystem functioning are rarely considered in the existing restoration programs. One current limitation to their use is the impossibility to determine an effective and site-specific list of species to be reintroduced for restoring soil chemical and physical properties. This can be solved by embracing the functional diversity of soil communities through adopting a functional trait-based approach. Instead of a classic taxonomic approach, functional traits can help to reduce context dependency and therefore enable generalization across communities and ecosystems. Pey et al. (2014) defined soil invertebrates’ traits as “morphological, physiological, phenological or behavioral features, measurable at the individual level, from the cell to the whole-organism level, without reference to the environment or any other level of organization”. While plant traits are already precisely incorporated in ecological restoration approaches (e.g. to select suitable plant species for restoration), this is rarely the case for soil organisms yet. Our working group of scientists in soil ecology and ecological engineering addressed the potential to use soil organisms’ functional traits to conceive sustainable and integrative ecosystem SUITMA restoration. We proposed a list of key traits related to three main ecological functions (soil structuration, nutrient and pollutant cycling process, biocontrol and preys’ regulation) that could be considered depending on the focus of the restoration plan. We further highlighted the main gaps of knowledge that need to be addressed for a successful integration of soil organisms’ functional traits in ecological engineering.

During the last few years, spoken language technologies have known a big improvement thanks to Deep Learning. However Deep Learning-based algorithms require amounts of data that are often difficult and costly to gather. Particularly, modeling the variability in speech of different speakers, different styles or different emotions with few data remains challenging. In this paper, we investigate how to leverage fine-tuning on a pre-trained Deep Learning-based TTS model to synthesize speech with a small dataset of another speaker. Then we investigate the possibility to adapt this model to have emotional TTS by fine-tuning the neutral TTS model with a small emotional dataset., Comment: Accepted at IntelliSys 2019

Les pelouses calcicoles de Haute-Normandie sont généralement considérées comme des formations herbacées semi-naturelles de grande valeur biologique. Consécutivement à leur abandon agricole depuis le début du XXe siècle, la biodiversité de ces milieux est menacée par une colonisation arborée spontanée. Une étude diachronique montre que cette dynamique est hétérogène. Les sites péri-urbains présentent différentes successions végétales marquées par une ancienne utilisation culturale. A l'inverse, les sites ruraux au passé plus pastoral, ont une évolution paysagère beaucoup plus homogène. Chalk grasslands of Normandy are generally considered as semi-natural grasslands with a high biological value. Consecutively to their agricultural abandonment since the beginning of the XXth century, the biodiversity of these ecosystem is threatened by a spontaneous colonisation by scrubs and forest. A diachronic study shows that this dynamic is heterogeneous. Peri-urbans sites present different vegetational-succession marked by a former cultural utilisation. In the opposite, rural-sites with pastoral past, have a more homogeneous landscape evolution.

La question fondamentale qui soutend toutes les recherches en dynamique des populations végétales peut s'exprimer ainsi : “pourquoi une espèce est-elle présente à tel moment et en telle proportion dans une communauté végétale et comment est-elle remplacée par d'autres lorsque certaines conditions environnementales changent ?”. Dans l'état actuel des connaissances en écologie végétale, il n'existe pas une théorie unifiée ou un modèle général pour répondre à cette question. Bien au contraire, f...

L'impact de différents régimes de gestion (fauchage, labour, abandon) sur la diversité végétale de quatre pelouses calcicoles de la vallée de Seine (Haute-Normandie, France) a été analysé par la méthode des points quadrats durant une période de 3 ans. Le classement des espèces en groupes végétaux fonctionnels permet de dégager les effets de la gestion conservatoire sur les phytocénoses indépendamment de la composition taxonomique des différentes pelouses. Les résultats des analyses multivariées montrent une dominance de la contribution des espèces annuelles et bisannuelles dans les quadrats régulièrement fauchés au contraire des quadrats abandonnés et des témoins plutôt dominés par des espèces pérennes. La richesse et la diversité spécifique des quadrats fauchés pendant trois ans sont significativement supérieures aux témoins et aux quadrats abandonnés. L’utilisation de groupes fonctionnels basés sur quelques caractéristiques des végétaux est discutée dans une optique de généralisation et de modélisation prédictive des effets de la gestion pour d'autres sites. The impact of different management regimes (mowing, ploughing, abandonment) on plant diversity of four chalk grasslands in the Seine valley (Upper-Normandy, France) have been analysed by the point quadrat method during a 3 year-period. Species classification by functional vegetation groups allows us to draw the effects of conservation management on phytocenosis independently of taxonomic composition of the different grasslands. Results of multivariate analysis show a dominance of annual and biennial species contribution in the regularly mown plots at contrary of controls and abandoned plots rather dominated by perennial species. Species richness and diversity of mown plots during three years are significatively higher than controls and abandoned plots. The use of functional groups based on several plant traits is discussed in a view of making predictive modélisation and generalisation of management effects to others sites.

L'étude de la macrofaune du sol au cours d'une succession secondaire potentielle a été réalisée dans différentes parcelles de pelouses calcicoles après abandon du pâturage. Les résultats montrent une évolution contrastée des différentes populations en rapport avec l'apparition d'une couverture arborée et des changements dans la qualité des litières. Dans les parcelles à végétation herbacée, les espèces épigées dominent contrairement aux parcelles boisées. Si la richesse taxonomique est maximum dans la pelouse en exclos, chaque parcelle demeure cependant caractérisée par la présence de certains groupes taxonomiques. En conséquence, les systèmes de gestion conservatoire mis en place pour la conservation biologique de pelouses calcicoles riches en plantes et en invertébrés doivent également préserver la diversité des communautés de macrofaune du sol par la réalisation de systèmes de gestion tournants. The study of soil macrofauna communities has been realized in different plots of chalk grasslands after the cessation of grazing along a potential secondary succession. Results show a contrasted evolution of the different populations in relation with the apparition of a tree cover and changes in litter quality. In plots with herbaceous vegetation, epigeous species dominate at the opposite of woody plots. If taxonomic richness is the highest in the enclosure grassland, each plot remains characterized by the presence of certain taxonomic groups. In consequence, conservation management systems defined for biological conservation issues in chalk grasslands (species-rich plant and invertebrate communities) must take into account the diversity of soil macrofauna communities with the realization of rotational management systems.

Les mécanismes des successions végétales sur une pelouse calcicole sont étudiés par les méthodes d'analyse diachronique de l'écologie historique. Les transformations observées sont (1) des variations dans la colonisation des parcelles et (2) la différenciation de deux types de pelouse, rase ou haute. Le déterminisme de ces contrastes est à rechercher dans les pratiques anciennes (effet rémanent du labour) et les stratégies compétitives ou disséminatrices des espèces végétales.

Les facteurs écologiques qui influencent la composition botanique et la structure de la végétation des pelouses calcaires de Normandie centrale sont appréhendés grâce à une étude régionale. Les résultats des analyses multivariées montrent que: (1) Deux types d’habitats peuvent être définis à partir des données climatiques et pédologiques (mésophile et mésoxérophile). (2) Les facteurs d’exploitation ont également une influence importante. Les processus d’intensification (accroissement de la pression de pâturage) sont ainsi responsables de la perte de richesse spécifique des pelouses calcicoles la plus rapide, tandis que l’abandon du pâturage conduit à des changements plus lents. En considérant les pelouses calcicoles dans leurs contextes agricoles, nous insistons alors sur la nécessité de procéder à une approche multi-niveaux pour répondre aux objectifs de la conservation biologique de ces milieux. From a regional survey carried out in central Normandy, we focuse on ecological factors which influence the botanical composition and vegetation structure of chalk grasslands. Results of the multivariate analysis show that: (1) Two majors types of habitats can be defined from climatic and pédologie data (mesophilous and mesoxerophilous). (2) Management factors have also notable influence on plant communities. Intensification processes (i.e. increasing grazing pressure) appear to be responsible of the more and rapid loss of the species-richness in chalk grasslands, while grazing abandonment leads sligher slower changes. Considering the chalk grasslands in their agricultural contexts, we then insist on the necessity of a multi-level approach for the conservation objectives.