Search

This paper proposes an instrumental variable approach for continuous-time system identification using fractional models with multiple input single output context. This work is an extension of the simplified refined instrumental variable approach (srivcf ) developed for single input-single output fractional model identification (Malti et al. (2008a); Victor et al. (2013)) to the multiple input-single output case. Monte Carlo simulation analysis is used to demonstrate the performance of the proposed approach. A study is then provided to motivate differentiation order estimation, and more specifically, commensurate order estimation.

Assessing the harmfulness of defects based on images is becoming more and more common in industry. Today these defects can be insert in digital twins that aim to replicated in a mechanical model what is observed on a component. We propose a methodology for defect classification and defect labeling in view of fast prediction of their harmfulness by using hyper-reduced order models. Mechanical models studied here are non parametric. This problematic issue is circumvent by introducing a classification of defect. Then, a local hyper-reduced order model is defined in each class of defect. The proposed methodology for the classification, consists in seeing each defect by its mechanical effects more than by its morphology. Each defect is encoded as a point on a Grassmann manifold. This data encoding uses a projection-based hyper-reduction. We restrict our attention to a mechanical modeling of elastic stresses around defects. We show that the hyper-reduced equations are similar to an oblique projection of the finite element prediction. Hence we obtain an upper bound of the approximation error for displacement that depends on a Chordal distance between points on a Grassmann manifold. From the available data set of defects, we have observed that the Grassmann distance is magnifying the Chordal distance between defects. A k-medoids algorithm is used in order to create clusters of defects from a matrix of Grassmann distance between defects. A simulation-based labeling of defect is proposed. This labeling procedure relies on hyper-reduction and a related error indicator. This error indicator measures the discrepancy between the stresses predicted via hyper-reduction and equilibrated stresses. In spite of the large variety of defects in the data set, we show accurate predictions of stresses for most of defects in the test set, while training the clustering and the hyper-reduced order model on a different training set. These accurate results are obtained by combining two contributions for a reduced basis for displacements : macroscopic modes related to a defect-free mechanical problem and elastic fluctuation modes specific to each class of defect. In this paper, we pay particular attention to the compromise to be made between the accuracy of the numerical approximations and the memory space required to save the reduced bases., L’évaluation de la nocivité des défauts à partir d’images est de plus en plus courante dans l’industrie. Aujourd’hui, ces défauts peuvent être insérés dans des jumeaux numériques qui visent à reproduire dans un modèle mécanique ce qui est observé sur un objet. Nous proposons une méthodologie pour la classification et l’étiquetage des défauts en vue de prédire rapidement leur nocivité en utilisant des modèles d’ordre hyperréduits. Les modèles mécaniques étudiés sont non paramétriques. Cette difficulté est contournée par l’introduction d’une classification des défauts. Ensuite, un modèle d’ordre hyperréduit local est défini dans chaque classe de défaut. La méthodologie proposée pour la classification, consiste à voir chaque défaut par ses effets mécaniques plus que par sa morphologie. Chaque défaut est encodé comme un point dans une variété de Grassmann. Ce codage des données utilise un modèle d’ordre réduit par projection. Nous nous limitons notre attention à une modélisation mécanique des contraintes élastiques autour des défauts. Nous montrons qu’une limite supérieure de l’erreur d’approximation pour le déplacement, dépend de la distance Chordal entre les points d’une variété de Grassmann, si une projection de Galerkin est effectuée pour la réduction du modèle. A partir de l’ensemble des données disponibles sur les défauts, nous avons observé que la distance de Grassmann amplifie la distance Chordal entre les défauts. Un algorithme de k-médoïdes est utilisé pour créer des groupes de défauts à partir d’une matrice de la distance de Grassmann entre les défauts. Un étiquetage des défauts basé sur la simulation est proposé. Cette procédure d’étiquetage repose sur l’hyper-réduction et un indicateur d’erreur associé. Cet indicateur d’erreur mesure l’écart entre les contraintes prédites par hyper-réduction et des contraintes équilibrées. Malgré la grande variété de défauts dans l’ensemble de données, nous montrons des prédictions précises des contraintes pour la plupart des défauts dans l’ensemble de test, tout en entraînant la classification et l’hyper-réduction sur un ensemble d’entraînement différent. Ces résultats précis sont obtenus en combinant deux contributions dans une base réduite pour les déplacements : des modes macroscopiques liés à un problème mécanique sans défaut et des modes de fluctuation élastique spécifiques à chaque classe de défaut. Dans cet article, nous accordons une attention particulière au compromis à faire entre la précision des approximations numériques et l’espace mémoire nécessaire pour enregistrer les bases réduites.

International audience; Belief propagation and variants are the usual decoding algorithms for low density parity check codes. In this paper, successive overrelaxations are applied to increase convergence speed of the decoding process. Performances are reported and improvements over reference algorithms are observed. This technique is also compatible with high speed parallel group decoding involved in modern standards like DVB-S2.

International audience; This paper proposes a new construction of rate-adaptive coding schemes based on Low Density Parity Check (LDPC) codes for Slepian Wolf source coding. Unlike standard rate-adaptive source coding schemes, the code construction we propose is based on finite-length code design tools that permit to greatly improve the decoding performance at short to moderate length. In particular, our method permits to limit the number of short cycles in the codes at all rates of interest, and to avoid eliminating some source bits from the code constraints. The proposed method shows a performance improvement of up to an order of magnitude at almost all the considered rates compared to the standard LDPCA construction.

International audience; Lowering the power supply of a circuit can induce transient errors in the memory cells and timing errors in the computation units. In this paper, we consider the Taylor-Kuznetsov (TK) memory architecture with transient errors in the memory cells and with timing errors in the correction circuit. We provide a theoretical analysis of the performance of TK memories under transient errors and timing errors. Our study is based on the analysis of the computation trees of the equivalent Gallager B decoders with and without timing errors. As a main result, we show that as the number of iterations goes to infinity, the error probability of the decoder with timing errors converges to the error probability of the decoder without timing errors. Monte Carlo simulations confirm this result even for moderate code lengths.

International audience; Abstract-In this paper, an efficient iterative receiver for digital communication systems is investigated. It combines a Non-Binary Low-Density Parity-Check (NB-LDPC) decoder with a high-order constellation demapper and a Multiple-Input Multiple-Output (MIMO) detector. A suboptimal MIMO detector based on the Belief Propagation (BP) algorithm is investigated as an alternative to a Maximum Likelihood (ML) detector. Extrinsic information is exchanged between the detector and the decoder thanks to an iterative process. EXtrinsic Information Transfer (EXIT) charts enable to analyze the convergence behaviour of the proposed MIMO-BP detector. A suitable schedule for the different types of iterations is finally proposed to reduce the receiver latency and improve its error correction performance.

National audience; Dans cet article, nous nous sommes intéressés aux problèmes de reconnaissance noncoopérative de cibles (NCTR) en tant que problème de classification supervisée. Après une présentation du système d'acquisition des profils distance radar et du problème de reconnaissance, suivie d'une étude statistique des données, nous proposons d'utiliser un algorithme des K plus proches voisins (KPPV) dont les performances sont détaillées en fonction du nombre de voisins K, du type de distance utilisée et de la nature des données utilisées (débruitées ou non). Dans un second temps, cet algorithme a été parallélisé sur un processeur many-coeurs (GPU : Graphics Processing Unit). Les opérations arithmétiques et le modèle d'accès mémoire ont été étudiés pour obtenir la meilleure parallélisation des calculs. Enfin, nous terminons par une discussion autour des perspectives envisageables pour la méthode proposée, notamment en s'intéressant à d'autres espaces de représentation ou à d'autres méthodes de classification. ABSTRACT. In this paper, first, we present the problem of Non Cooperative Target Recognition (NCTR) as a supervised classification problem. After a presentation on the radar acquisition system of range profiles and the problem of recognition, followed by a statistical study of data, we use a classical classification method of K Nearest Neighbors (KNN) to do this classification. We explore and compare the performances of this algorithm based on the choice of the distances, the choice of K and the nature of used data (denoised or not). KNN algorithm has been executed initially on CPU with Matlab and then on GPU. Arithmetic operations and memory access pattern has been studied to get the best parallelization. Finally, we conclude with a discussion about possible perspectives for the proposed method especially by focusing on other representation spaces or other classification methods.

International audience; Branch Prediction is a widely used technique to optimize pipelined microprocessor architectures. In this paper, a High Level Synthesis (HLS) approach combining operation speculation and branch prediction is presented. In the proposed design flow, the CDFG (Control Data Flow Graph), is obtained by compiling the application. A speculation Graph is built. This graph allows to evaluate the potential of each branch to be predicted and the potential of each related Basic Block BB to be speculated. A target BB is then selected and scheduled using a list scheduling algorithm. A couple of BBs that will be predicted during the execution of the target BB is created. Operations of the couple are associated and speculativally scheduled in the target BB before being bounded to functional units. This step is repeated until all the BBs are scheduled. Finally, the RTL architecture is generated. The interest of branch prediction techniques in term of timing performances, and its impact on area are evaluated through first experiments.

The probability hypothesis density (PHD) methodology is widely used by the research community for the purposes of multiple object tracking. This problem consists in the recursive state estimation of several targets by using the information coming from an observation process. The purpose of this paper is to investigate the potential of the PHD filters for real-time traffic state estimation. This investigation is based on a Cell Transmission Model (CTM) coupled with the PHD filter. It brings a novel tool to the state estimation problem and allows to estimate the densities in traffic networks in the presence of measurement origin uncertainty, detection uncertainty and noises. In this work, we compare the PHD filter performance with a particle filter (PF), both taking into account the measurement origin uncertainty and show that they can provide high accuracy in a traffic setting and real-time computational costs. The PHD filtering framework opens new research avenues and has the abilities to solve challenging problems of vehicular networks.

International audience; The (24, 12, 8) extended binary Golay code is a well-known rate-1/2 short block-length linear error-correcting code with remarkable properties. This paper investigates the design of an efficient low-complexity soft-decision decoding architecture for this code. A dedicated algorithm is introduced that takes advantage of the code’s properties to simplify the decoding process. Simulation results show that the proposed algorithm achieves close to maximum-likelihood performance with low computational cost. The decoder architecture is described, and VLSI synthesis results are presented.

International audience; In this paper we address the joint interpolation of missing data and estimation of ocean surface velocities from multi-resolution sea surface satellite observations. A variational assimilation model is proposed. Using synthetic simulation and real SST data, we conducted experiments to evaluate the relevance of the proposed model, in particular the relevance of the fusion of observations at different resolutions and the improvement issued from the consideration of dynamics prior in the assimilation model. Numerical and qualitative results assessed the effectiveness of the proposed methods.

LGEP 2011 ID = 739; International audience; This paper shows that Ridge Polynomial Neural Networks (RPNN) and Least-Square Support Vector Machines (LS-SVM) technique provide efficient tools for microwave characterization of dielectric materials. Such methods avoids the slow learning properties of multilayer perceptrons (MLP) which utilize computationally intensive training algorithms and can get trapped in local minima. RPNN and LS-SVM are combined in this work with the Finite Element Method (FEM) to evaluate the dielectric materials properties. The RPNN is constructed from a number of increasing orders of Pi-Sigma units, it maintains fast learning properties and powerful mapping capabilities of single layer High Order Neural Networks (HONN). LS-SVM is a statistical learning method that has good generalization capability and learning performance. The FEM is used to create the data set required to train the RPNN and LS-SVM. The performance of a LS-SVM model depends on a careful setting of its associated hyper-parameters. In this study the LS-SVM hyper-parameters are optimized by using a Bayesian regularization technique. Results show that LS-SVM can achieve good accuracy and faster speed than neural network methods.

The design process of complex mechanical systems is composed of the embodiment design phase, where the main structuring characteristics are chosen. Constraints on discrete and continuous variables are used to model such problems. The embodiment design phase is analysed and some relevant vaiable classes are highlighted and illustrated with an heat batchexchanger. Some specific search heuristics applied during the resolution phase are define in this paper.

International audience; This paper presents a simple approach to assess performance that can be achieved by the MMSE turbo equalizer in ST-BICM systems over multipaths Rayleigh block fading channels with i.i.d fading statistics. By considering perfect information exchange between the SISO decoder and the linear MMSE equalizer, the performance reaches the matched filter bound, and a closed form expression of the corresponding probability of bit error can be derived at the output of the equalizer. In particular, we emphasize that the suggested approach provides an attractive and reliable tool for performance validation consistently with the proposed expression of the probability of bit error. Simulations for 4-PSK and 8-PSK modulated signals show the relevance of the proposed approach and the full benefit provided by the MMSE turbo equalizer. In addition some clarification of the signal-to- noise ratio definition is pointed out.

International audience; In this paper, we present an original use of Reed-Solomon (RS) codes as space-time block codes (STBCs) and compare them to orthogonal STBCs (OSTBCs). Due to their symbol structure, they can have a representation close to that of an STBC. RS codes are admittedly more complex to decode as they have no orthogonal properties. However, they offer much higher coding rates. In addition, a wide range of these codes can be easily generated. To decode them, we have developed a new algorithm using a sliding encoding-window (SEW), which is combined with the decoding-based Chase algorithm. We compare the RS(15,7) code to the Alamouti one, both having an equivalent coding rate. Its performance compared to Alamouti gets better as the receive diversity increases. Moreover, RS codes do not require the channel to remain constant over the code duration. In fact, they offer better results if the channel changes rapidly, which can be obtained virtually by introducing interleaving

International audience; Interleaving is a key component of many digital communication systems involving error correction schemes. It provides a form of time diversity to guard against bursts of errors. Recentlty, interleavers have become an even more integral part of the code design itself, if we consider for example turbo and turbo-like codes. In a non-cooperative context, such as passive listening, it is a challenging problem to estimate the interleaver parameters. In this paper, we propose an algorithm that allows us to estimate the parameters of the interleaver at the output of a binary symmetric channel and to locate the code-words in the interleaved block. This gives us some clues about the interleaving function used.

In object-oriented or object-relational databases such as multimedia databases or most XML databases, access patterns are not static, i.e., applications do not always access the same objects in the same order repeatedly. However, this has been the way these databases and associated optimisation techniques such as clustering have been evaluated up to now. This paper opens up research regarding this issue by proposing a dynamic object evaluation framework (DOEF). DOEF accomplishes access pattern change by defining configurable styles of change. It is a preliminary prototype that has been designed to be open and fully extensible. Though originally designed for the object-oriented model, it can also be used within the object-relational model with few adaptations. Furthermore, new access pattern change models can be added too. To illustrate the capabilities of DOEF, we conducted two different sets of experiments. In the first set of experiments, we used DOEF to compare the performances of four state of the art dynamic clustering algorithms. The results show that DOEF is effective at determining the adaptability of each dynamic clustering algorithm to changes in access pattern. They also led us to conclude that dynamic clustering algorithms can cope with moderate levels of access pattern change, but that performance rapidly degrades to be worse than no clustering when vigorous styles of access pattern change are applied. In the second set of experiments, we used DOEF to compare the performance of two different object stores: Platypus and SHORE. The use of DOEF exposed the poor swapping performance of Platypus., Comment: 20 pages

International audience; This paper provides insight INTO data_ds.tmp_pub_the statistical distribution of the interference at the output of the receivers composed of a chip-rate linear transverse equalizer followed by a despreader for the downlink of direct-sequence code-division multiple access (DS-CDMA) systems with long scrambling sequences. It is shown that for small spreading factors, the probability density function (PDF) of the interference samples can be modelled accurately as a mixture of Gaussian PDFs. Each component of the mixture is associated with a particular realization of the scrambling sequence spanning the spread symbol to be retrieved. Based on the study, an analytical method allowing the computation of the bit error probability for small spreading factors is derived. Numerical simulations show that the BEP estimation achieved by the proposed method is far more accurate than that obtained via the traditionnal Gaussian assumption on the interference statistics.

This paper presents a block turbo decoding algorithm, from its theory to its implementation in a programmable circuit. In this study, we discuss the two prototypes realized. It will be possible to compare the complexity of the core of the process, which is the elementary decoder, thanks to the choice of essential parameters. One prototype is more dedicated to high data rates, the other one being implemented on only one FPGA which means a gain in terms of area. First, we briefly focus on the description of the siso (Soft-In Soft-out) algorithm used to implement the turbo decoder. Then, we explain the essential choices in order to adapt the algorithm for an ASIC implementation, which leads to a compromise between area and binary error rate. Finally, we present the two prototypes implemented and their experimental results. L’objectif de cet article est de presenter le turbodecodage de codes produits depuis la theorie jusqu’a la realisation sur circuit programmable. Deux prototypes ont ete realises, Us serviront de support a cette etude. Il sera en effet possible de comparer la complexite du decodeur elementaire en fonction d’un certain nombre de parametres, de mettre face a face une structure dediee aux debits eleves et une structure optimisee en surface. Dans une premiere partie, on trouvera une breve description de l’algorithme srso (Soft-In Soft-Out) permettant de realiser le turbo decodage. Vient ensuite la description de l’adaptation de cet algorithme a l’implantation sur silicium. Cette phase necessite des choix entre complexite en terme de circuit et performance en terme de taux d’erreurs binaires. La derniere partie presente les choix d’architectures et les resultats experimentaux des 2 prototypes.

International audience; In this paper we propose an optimized transmission system for still images. This system comprises a VQ source encoder with a compression rate of 9. The source data is encoded using a BCH product code and mapped on a 16-QAM constellation. The codebook and channel encoder have been optimized to minimize distortion in presence of transmission errors. We use a new joint source-channel turbo decoding algorithm of the product code. As compared to classical channel decoding algorithms (soft decoding), we obtain an improvement of 0.7 dB in terms of coding gain for a BER of 10/sup -5/. The PSNR of the received image exhibits an improvement of 11 dB as compared to conventional turbo decoding at low SNRs.

In this paper, we proposed a approach to the two-microphone speech enhancement problem based on the Blind Separation of Sources for convolutive mixtures. The separation of signals can be achieved by a recursive structure of adaptive Filters FIR. The Filter coefficients must be updated by using Higher Order Statistics in order to insure the output independence. Various adaptive algorithms based on different independence criteria are proposed to estimate the filters. The experimental results have been obtained for different types of signals in various situations of mixtures. For instance, using two microphones, the algorithms are able to separate two speech signals without any knowledge of the signals