Search

Cet article présente la caractérisation du comportement d'un moniteur de mesure de la contamination particulaire radioactive atmosphérique en condition de chantier de démantèlement simulée en laboratoire. Les premiers résultats présentés dans ce papier mettent en avant une mauvaise adaptation de la compensation dynamique du bruit de fond dans des conditions de fonctionnement non prises en compte dans les référentiels IEC., This paper presents the characterization of the behavior of a Continuous Air Monitor (CAM) under dismantling site conditions simulated in laboratory. The first results presented in this paper highlight a misinterpretation of the dynamic background compensation in operating conditions not taken into account in IEC standards.

Ce mémoire est composé de trois articles qui s’unissent sous le thème de la recommandation musicale à grande échelle. Nous présentons d’abord une méthode pour effectuer des recommandations musicales en récoltant des étiquettes (tags) décrivant les items et en utilisant cette aura textuelle pour déterminer leur similarité. En plus d’effectuer des recommandations qui sont transparentes et personnalisables, notre méthode, basée sur le contenu, n’est pas victime des problèmes dont souffrent les systèmes de filtrage collaboratif, comme le problème du démarrage à froid (cold start problem). Nous présentons ensuite un algorithme d’apprentissage automatique qui applique des étiquettes à des chansons à partir d’attributs extraits de leur fichier audio. L’ensemble de données que nous utilisons est construit à partir d’une très grande quantité de données sociales provenant du site Last.fm. Nous présentons finalement un algorithme de génération automatique de liste d’écoute personnalisable qui apprend un espace de similarité musical à partir d’attributs audio extraits de chansons jouées dans des listes d’écoute de stations de radio commerciale. En plus d’utiliser cet espace de similarité, notre système prend aussi en compte un nuage d’étiquettes que l’utilisateur est en mesure de manipuler, ce qui lui permet de décrire de manière abstraite la sorte de musique qu’il désire écouter., This thesis is composed of three papers which unite under the general theme of large-scale music recommendation. The first paper presents a recommendation technique that works by collecting text descriptions of items and using this textual aura to compute the similarity between them using techniques drawn from information retrieval. We show how this representation can be used to explain the similarities between items using terms from the textual aura and further how it can be used to steer the recommender. Because our system is content-based, it is not victim of the usual problems associated with collaborative filtering recommenders like the cold start problem. The second paper presents a machine learning model which automatically applies tags to music. The model uses features extracted from the audio files and was trained on a very large data set constructed with social data from the online community Last.fm. The third paper presents an approach to generating steerable playlists. We first demonstrate a method for learning song transition probabilities from audio features extracted from songs played in professional radio station playlists. We then show that by using this learnt similarity function as a prior, we are able to generate steerable playlists by choosing the next song to play not simply based on that prior, but on a tag cloud that the user is able to manipulate to express the high-level characteristics of the music he wishes to listen to.

cote interne IRCAM: Warusfel04a; None / None; National audience; Future generation of audio content will promote immersion and presence, both notions strongly related to the spatial attributes of the auditory scene. Presence is particularly stimulated in active situations where the listener may navigate in the soundscape or interact with the sound objects. In such situations, enabled by WFS or binaural rendering, the continuous update and congruence of the acoustical cues, depending on listener's actions, have a strong impact on presence sensation. Among these acoustical cues, the paper focuses on directional properties of sound source radiation. The paper first reviews the perceptual and cognitive aspects involved by the radiation properties of sound sources and illustrates them in the context of future audio applications. These applications range from the design of various audio devices or active material with controllable radiation patterns, to the sound field reproduction with accurate rendering of directivity properties. In the last example, the stake is to render convincing 3D sound objects instead of 2D sound pictures, allowing the listener to experience coherent direct sound contributions and room effect while wandering around the soundscape. The paper describes different methods dedicated to the reproduction of sound source directivity. They are based on a physical modeling of the radiation properties and use array signal processing. A first approach is studied where a 3D array of transducers is located on stage and is controlled in order to radiate sound in a way equivalent to the simulated source. A dual approach, where the transducer array is distributed around the audience uses WFS in order to create a sound field similar to the one that would have been generated by the virtual source. These two approaches are compared, in their respective application domains, in terms of reproduction accuracy of the direct sound and associated room effect, joined recording technique, practical implementation and flexibility.

The thesis work presented in this paper aims to address the issues encountered in the field of decommissioning and related to non-destructive testing of large non-planar surfaces with low levels of radiological contamination. The main objective of this thesis is to develop a pixelated and deformable measurement system, compatible with the detection of different radiological signatures (alpha, beta and gamma) on large uneven surfaces. Two technical solutions were considered and investigated: scintillating fibres and phoswichs scintillators. Scintillating fibres, being characterised by an intrinsic deformability allowing them to adapt to the measured surface and the possibility of being manufactured over several kilometres in length, showed a strong potential for innovation and therefore constituted the most promising technological brick for this work. In order to achieve the stated objective, three lines of research were explored. The first axis concerns the development of a detector based on scintillating fibres, for a beta spectrometry application. The second axis aims at developing a numerical model of the detector, using the Monte Carlo simulation code MCNP6, and a numerical model of the transmission and amplification of the signal generated inside the fibres. Finally, the third axis concerns a multi-source beta/beta and beta/gamma discrimination method based on different thicknesses of cladding and cores, and associated with an iterative Maximum Likelihood-Expectation Maximisation (ML-EM) algorithm.; Les travaux de thèse présentés dans ce mémoire visent à répondre aux enjeux rencontrés dans le domaine du démantèlement et relatifs aux contrôles non destructifs de grandes surfaces non planes, présentant un faible niveau de contamination radiologique. L’objectif principal de la thèse consiste à développer un système de mesure pixellisé et déformable, compatible avec la détection de différentes signatures radiologiques (alpha, bêta et gamma) et adapté à l’inspection de grandes surfaces non planes. Deux solutions techniques ont été considérées et investiguées : les fibres scintillantes et les scintillateurs de type phoswich. Les fibres scintillantes ont montré un fort potentiel d’innovation et ont donc constitué la brique technologique la plus prometteuse pour ces travaux.Afin d’atteindre l’objectif énoncé, trois axes de recherche ont été explorés. Le premier axe concerne le développement d’un détecteur basé sur des fibres scintillantes, pour une application de spectrométrie bêta. Le deuxième axe a pour objectif le développement d’un modèle numérique du détecteur, en utilisant le code de simulation Monte-Carlo MCNP6 et un modèle mathématique de la transmission et de l’amplification du signal généré à l’intérieur des fibres. Enfin, le troisième axe porte sur une méthode de discrimination multi-sources bêta/bêta et bêta/gamma basée sur différentes épaisseurs de gaines et de cœurs, et associée à un algorithme itératif de type Maximum Likelihood-Expectation Maximisation (ML-EM).

National audience; In this paper, we propose an efficient method to estimate in a nonparametric fashion the marks' density of a shot-noise process in presence of pileup from a sample of low-frequency observations. Based on a functional equation linking the marks' density to the characteristic function of the observations and its derivative, we propose a new time-efficient method using B-splines to estimate the density of the underlying gamma-ray spectrum, which is able to handle large datasets used in nuclear physics. A discussion on the numerical computation of the algorithm and its performances on simulated data are provided to support our findings.; Nous proposons une méthode d'estimation non-paramétrique rapide pour estimer la distribution des marques d'un processus de shot-noise en présence d'empilement à partir d'un nombre potentiellement important d'observations mais échantillonnées à basse fréquence. À partir d'une équation fonctionnelle liant la densité des marques à la fonction caractéristique des observations et sa dérivée, nous proposons un estimateur de cette densité en utilisant la base des B-splines. Nous discutons de l'implémentation pratique de l'algorithme et illustrons les performances de l'estimateur sur des données simulées.

In this paper, we study an inverse scattering problem on Liouville surfaces having two asymptotically hyperbolic ends. The main property of Liouville surfaces consists in the complete separability of the Hamilton-Jacobi equations for the geodesic flow. An important related consequence is the fact that the stationary wave equation can be separated into a system of a radial and angular ODEs. The full scattering matrix at fixed energy associated to a scalar wave equation on asymptotically hyperbolic Liouville surfaces can be thus simplified by considering its restrictions onto the generalized harmonics corresponding to the angular separated ODE. The resulting partial scattering matrices consists in a countable set of $2 \times 2$ matrices whose coefficients are the so called transmission and reflection coefficients. It is shown that the reflection coefficients are nothing but generalized Weyl-Titchmarsh functions for the radial ODE in which the generalized angular momentum is seen as the spectral parameter. Using the Complex Angular Momentum method and recent results on 1D inverse problem from generalized Weyl-Titchmarsh functions, we show that the knowledge of the reflection operators at a fixed non zero energy is enough to determine uniquely the metric of the asymptotically hyperbolic Liouville surface under consideration., 39 pp

National audience; This paper addresses the problem of computing the optimal bayesian estimator (in the mean square sense) in some jump Markov models. We make use of a recent algorithm which computes this estimator fastly and exactly in some particular Triplet Markov models. The problem then reduces to a parameter estimation problem; our two-step estimator is based on a Kullback-Leibler divergence minimization and an unsupervised parameter estimation step; Cette communication traite du problème du calcul de l'estimateur bayésien optimal (au sens de l'erreur quadratique moyenne) dans des modèles markoviens à sauts. Nous utilisons pour ce faire une technique récente de calcul rapide et exact de cet estimateur dans une classe de modèles de Markov triplet particuliers. Le problème se ramène alors au problème d'estimation des paramètres de ces modèles; nous proposons une méthode d'estimation basée d'une part sur la minimisation d'une divergence de Kullback-Leibler et d'autre part sur une étape d'estimation non supervisée

The microstate model describes EEG signals as series of topographies remaining stable during several tens of milliseconds and associated to brain states. The proposed generalization in this paper is based on an overcomplete model allowing several states to be active simultaneously. A dictionary learning algorithm with a temporal regularization is proposed to extract these states. The representation effectiveness of both models is compared on artificial and real signals for the extraction of the evoked potential P300., Le modèle des micro-états décrit les signaux EEG par des suites de topographies associées à des états cérébraux demeurant stables durant quelques dizaines de millisecondes. La généralisation proposée dans cet article considère un modèle redondant autorisant plusieurs états à être actifs simultanément. Un apprentissage de dictionnaire régularisé temporellement est proposé afin d'extraire ces états. L'efficacité de représentation des deux modèles est comparée sur des signaux de synthèse ainsi que sur des signaux réels pour l'étude du potentiel évoqué P300.

This paper focuses on spatially regularized overcomplete dictionary decompositions for EEG signals. We compare different combinations of parcimonious and spatial regularizations leading the decomposition to solutions physiologically plausible. The constructed multichannel EEG-oriented model is formalized in a convex framework and the non-differentiable decomposition problem is solved using a proximal iterative approach. The behaviors of the introduced priors for the denoising of EEG signal are assessed both on synthetic and real EEG data where our approach is considered for P300 single trial classification., Ce papier considère la décomposition parcimonieuse de signaux EEG sur un dictionnaire sur-complet temps fréquence.Nous comparons ici différentes combinaisons de régularisations parcimonieuses et spatiales guidant la décompositionvers des solutions plausibles physiologiquement. Notre approche est formalisée dans un problème d’optimisation convexenon-différentiable résolu grâce à une méthode proximale. L’efficacité de ces régularisations pour le débruitage de signaux EEGest testée à la fois sur des données artificielles et sur des données réelles, où notre approche est utilisée pour l’identification depotentiels évoqués P300.

National audience; The rapid changes in embedded electronics these recent years are the result of permanent improvements of design methodologies in adequacy with constraints that support the application. In Software Defined Radio field, the electronic system has the property of being "Real-Time" because it must respond within a specific time interval, known in advance. The growing need for software radio equipment with the ability to automatically adapt to the environment in which it is located, leads to the simultaneous execution of a large number of software components on a single processor in future broadband software radio. Consequently, these radios equipments become highly time constrained. From this perspective, current design methodologies for software radio integrating real-time scheduling analysis cannot be applied. This paper presents a new MDE (Model Driven Engineering) design methodology, allowing specifying parameters, which make possible real-time scheduling analysis of software components executing on a uniprocessor in a software radio. Also, we propose an algorithm, implemented as a tool, which simulates real-time scheduling of tasks implementing flexible signal processing algorithms. This tool is integrated to the proposed design methodology. Finally, we present an experiment of our methods on a project to demonstrate the benefits of our methodology and the good accuracy of our tool.

cote interne IRCAM: Lanchantin11a; None / None; National audience; Hidden Markov chains (HMC) are a very powerful tool in hidden data restoration and are currently used to solve a wide range of problems. However, when these data are not stationary, estimating the parameters, which are required for unsupervised processing, poses a problem. Moreover, taking into account correlated non-Gaussian noise is difficult without model approximations. The aim of this paper is to propose a simultaneous solution to both of these problems using triplet Markov chains (TMC) and copulas. The interest of the proposed models and related processing is validated by different experiments some of which are related to semi-supervised and unsupervised image segmentation.

International audience; In this paper, we develop two main results. The first one is a theorem proving that a Second Order Wiener Model can be blindly identified, i.e. using only the mean, the third and fourth order cumulants of the output data. The second result is the application of this theorem to spectral inversion (i.e. the recovering of the Power Spectrum Density) of the input signal of a Second Order Volterra Model to which usual inversion schemes cannot be applied, in particular when the linear kernel has a strong attenuation in frequency range. Numerical results are discussed with respect to the nonlinear energy amount of the output, the time series length and the SNR values.

cote interne IRCAM: Lanchantin08b; None / None; National audience; The hidden Markov chain (HMC) model is a couple of random sequences (X,Y), in which X is an unobservable Markov chain, and Y is its observable noisy version. Classically, the distribution p(y|x) is simple enough to ensure the Markovianity of p(x|y), that enables one to use different Bayesian restoration techniques. HMC model has recently been extended to "pairwise Markov chain" (PMC) model, in which one directly assumes the Markovianity of the pair Z=(X,Y), and which still enables one to recover X from Y. Finally, PMC has been extended to "triplet Markov chain" (TMC) model, which is obtained by adding a third chain U and considering the Markovianity of the triplet T=(X,U,Y). When U is not too complex, X can still be recovered from Y. Then U can model different situations, like non-stationarity or semi-Markovianity of (X,Y). Otherwise, PMC and TMC have been extended to pairwise "partially" Markov chains (PPMC) and triplet "partially" Markov chains (TPMC), respectively. In a PPMC Z=(X,Y) the distribution p(x|y) is a Markov distribution, but p(y|x) may not be and, similarly, in a TPMC T=(X,U,Y) the distribution p(x,u|y) is a Markov distribution, but p(y|x,u) may not be. However, both PPMC and TPMC can enable one to recover X from Y, and TPMC include different long-memory noises. The aim of this paper is to show how a particular Gaussian TPMC can be used to segment a discrete signal hidden with long-memory noise. An original parameter estimation method, based on "Iterative Conditional Estimation" (ICE) principle, is proposed and some experiments concerned with unsupervised segmentation are provided. The particular unsupervised segmentation method used in experiments can also be seen as identification of different stationarities in fractional Brownian noise, which is widely used in different problems in telecommunications, economics, finance, or hydrology.

In this paper we present a method for estimating the statistical properties of two well-known edge detectors: the non maxima suppression and the zero crossing of the Laplacian algorithms. Assuming the data are corrupted by an additive Gaussian noise we derive the probability density function (pdf) of the detected edge. Thanks to this approach the computed pdf explicitly depends on the parameters of the edge detector. Experimental results on real images and comparisons with Monte Carlo simulations are presented in order to characterize the performance of this method.

cote interne IRCAM: Schwarz07a; None / None; National audience; Corpus-based concatenative methods for musical sound synthesis have attracted much attention recently. They make use of a variety of sound snippets in a database to assemble a desired sound or phrase according to a target specification given in sound descriptors or by an example sound. With ever-larger sound databases easily available, together with a pertinent description of their contents, they are increasingly used for composition, high-level instrument synthesis, interactive exploration of a sound corpus, and others. This paper gives an overview of the components needed for corpus-based concatenative synthesis, and details of some realisations. Signal processing methods are crucial for all parts of analysis, (segmentation and descriptor analysis), for synthesis, and can intervene in the selection part e.g. for spectral matching.

National audience; This review paper deals with the use of on board full waveform LiDAR systems for monitoring continental surfaces. The objectives are 1/ to give an outline of the nature of information recorded by these systems, 2/ to present the main data processing methods used to process data acquired by both large and small footprint LiDAR systems and 3/ to evaluate the quality of ground and over ground characterisation obtained by processing LiDAR waveforms. The measurement process is first described and the main existing full waveform systems are presented. A second part deals with waveform processing methods: data visualization, deconvolution, extraction of geometrical and thematic information. Finally the main kinds of methods used to provide ground and over ground characterization from 3D point cloud and additional information derived from the waveforms are mentioned (DTM computation, building extraction, forest stand characterization. Research conducted in this field show that full waveform LiDAR data are of real interest for many applications: urban cartography, infrastructure and forest monitoring; Cet article, basé sur une analyse bibliographique, présente les systèmes Lidar embarqués de type full waveform et montre leur potentiel pour l'étude du sol et du sursol. Les objectifs sont 1/ de donner un aperçu de la nature de l'information enregistrée par ces systèmes, 2/ de présenter les principales méthodes de traitement des données acquises par les systèmes à large ou petite empreinte et 3/ d'évaluer la qualité de l'information extraite pour caractériser le sol et le sursol. Dans un premier temps le processus de mesure est décrit et les principaux systèmes full waveform existants sont présentés. Une seconde partie concerne les méthodes de traitement des formes d'onde LiDAR : visualisation des données, déconvolution, extraction d'informations géométriques et thématiques. Enfin les grandes familles de méthodes permettant de caractériser le sol et le sursol à partir des nuages de points 3D et des informations additionnelles dérivés du signal LiDAR sont évoquées : calcul de MNT extraction de bâtiments, description de la végétation forestière Les résultats des recherches menées dans ce domaine montrent que les données LiDAR full waveform sont pertinentes pour fournir de l'information dans de nombreux domaines d'application tels que la cartographie en milieu urbain, le suivi d'infrastructures, la caractérisation et le suivi de peuplements forestiers...

cote interne IRCAM: Rioux00a; None / None; National audience; The transient sound from an organ pipe is a very important component in deciding the pipe's timbre. This paper addresses the issue of verbal description of transient sounds of flue organ pipes. This problem must be solved when, as in our case, one wishes to document, understand and possibly assist the work of an organ-builder in the process of voicing organ pipes. Therefore, we explain the procedure used to map the relationship between verbal descriptors and quantitative acoustic analyses. The study is restricted to single flue organ pipes (no chords). A procedure integrating both objective and subjective methods is presented in order to develop such a description. After an introduction on basic voicing principles, a list of verbal descriptors used by the voicer is presented and serves as a basis for listening tests and signal analysis. Several types of listening tests are described. Computer implementation of these tests is shown to be particularly useful. Signal analysis using time-frequency methods is employed for separating harmonics and for representing the transient components of the signal. Results are also reported on the perception of the speed of the transient and of the sound “hiss”.

cote interne IRCAM: Tassart96a; /; National audience; Many studies have been undertaken on the modeling of physical systems by means of waveguidefilters. These methods consist mainly in simulating the propagation of acoustic waves with digitaldelay lines. These models are constrained to have a spatial step determined by the sampling ratewhich is a serious drawback when a high spatial resolution in the geometry of the model is neededor when the length of the waveguide needs to vary. One can use digital filters for approximatingthe exact fractional delay, but length variations usually induce audible distortions because of localinstabilities or modification of the filter's structure. Lagrange Interpolation theory leads to FIR filters which approximate fractional delays accordingto a maximally flat error criterion. Major drawbacks of current implementations of LagrangeInterpolator Filters (LIF) are a high computation cost and a lack of control over the delay whichcan only vary in a narrow range of values. We propose a new implementation of LIF based on a formal power series expansion of the exactz-transform. We have developed different fast and modular algorithms for LIF which make theLIF usable for real-time delay-varying applications. Modularity in the structure is a key pointhere as it enables one to switch between filters of different order while preserving the continuityof the z-transform. Thus the delay may vary over an unlimited range of values. Furthermore, anyarbitrary integer part of the fractional delay can be simulated by a classical delay line so that theactual order of the LIF may be maintained within reasonable limits. This paper will focus on thetime-varying properties of our implementation and its numerical stability over a wide range ofdelays.

In spectral analysis and pattern recognition, especially in speech processing, AutoRegressive (AR) modeling is widely used as a signal parameter estimation tool. In this paper, AR estimation behaviour is studied when applied to signals presenting abrupt changes at an unknown instant of time. A particular abrupt change is studied: the additive change case. This change occurs when, at the instant of change, a signal adds to one which was present before - both signals are independent -. The theoretical sliding window AR parameter expressions are given. We show that blind sliding window AR estimation can lead to a detection tool, allowing additive changes to be detected and distinguished

This paper describes the detection of stop consonants in the French language with the help of the wavelet transform. Our hypothesis is that the burst of stopconsonants can be approximated with a pulse. The detection system is based on the study of the correlation functions between the modulus of the impulse wavelet transform and the same transform of the speech signal. As we are using a Gaussian wavelet, the correlation functions have a minimum before each maximum, where that maximum is synchronous with the burst of the stop consonant. Two signals S M and S m are computed. The are respectively the synchronous and asynchronous sums of all the correlation functions. To localize the local minima of the S m function, two criteria are used. The temporal localisation of the pulse is accomplished with certain local maxima of the signal S M . The method of detection is tested with a corpus which has 137 unvoiced and 55 voiced stop consonants. The rate of detection is 94% for the unvoiced stops and 75% for the voiced stops. This study shows that for detection, the burst of stop consonants can be modelled with a pulse

The signals with an instantaneous frequency following an hyperbolic low are common in many fields of the signal processing, and particularly in the study of biologicals SONARS. In this paper we describe an harmonic analysis adapted to this kind of signals. The problem of stationnarity of stochastic processes will be adressed within the frame of the previous harmonic analysis. This allow us to define a power spectral density for second-order stationnary processes using the Bochner theorem. Some simulation results are presented showing the interest of the adapted harmonic analysis

The measurement of the velocity of propagation of ultrasonic pulse in dispersive media gives a quantity which physical meaning is ambiguous. In this paper we simulate a dispersive medium using the Kramers-Kronig relationship which expresses the causality of a linear system. We develope an appropriate signal processing such as Hilbert transform and cross-correlation to measure the time of flight of an ultrasonic impulse propagating in such a medium. This allows to deduce from this simulation the physical significance of the two informations given by two methods. Then these two methods are applied to several categories of graphites with dispersif character. Thus the differences between the measured velocities by Hilbert transform and cross-correlation are correlated to several microstructural properties of the material such as grain size, porosity and to thermal elaboration treatement

Developments in digital signal processing allow sound engineers to restore old recordings. The elementary tools used in the mentioned processes are digital editing of sound supported by the advanced visualization of the processed signal, equalizing filtering and adaptive digital filtering used to diminish continuous noise. Moreover, various methods of the elimination of impulse noises were tried. Some of the above methods were practically applied to the constructed computer sound restoration system presented in the paper. Consequently, the multi-task sound restoration method was proposed and practically employed. Application of some types of procedures to the sound restoration of old recordings will be discussed. The reduction of background noise using digital filtering in the critical bands was investigated. Another procedure for old magnetic tape echoes cancellation was implemented and tested. Application of smoothing filters to the elimination of noises will be also discussed

The work presented in this paper deals with the decomposition of pseudoperiodic signals in periodic components by using a spectral interpolation method. This method is applied to the autocorrelation of a pseudoperiodic signal instead of the signal itself. Results obtained by the present method are not very sensitive to the noise effect and are generally more accurate than those directly obtained by using the spectral interpolation method on the signal

This paper is an experimental investigation of the aerodynamic noise sources in a centrifugal fan with outlet guide vanes and axial outflow. Special piezoplastic sensors are used for the measurement of the pressure fluctuations on the blades. Classical signal processing procedures provide the blade loadings harmonics. When the fan operates at a relatively high flow rate and low efficiency, it is verified that higher loading harmonics are responsible for the noise emitted at the blade passing frequency. Conversely, at the maximum efficiency, only the first loading harmonic is noticeable, and the tone noise is suppressed