International Professional University of Technology in Nagoya Repository
Not a member yet
15131 research outputs found
Sort by
Rank-1 Constrained Multichannel Wiener Filter for Speech Recognition in Noisy Environments
No full textInternational audienceMultichannel linear filters, such as the Multichannel Wiener Filter (MWF) and the Generalized Eigenvalue (GEV) beamformer are popular signal processing techniques which can improve speech recognition performance. In this paper, we present an experimental study on these linear filters in a specific speech recognition task, namely the CHiME-4 challenge, which features real recordings in multiple noisy environments. Specifically, the rank-1 MWF is employed for noise reduction and a new constant residual noise power constraint is derived which enhances the recognition performance. To fulfill the underlying rank-1 assumption , the speech covariance matrix is reconstructed based on eigenvectors or generalized eigenvectors. Then the rank-1 constrained MWF is evaluated with alternative multichannel linear filters under the same framework, which involves a Bidirectional Long Short-Term Memory (BLSTM) network for mask estimation. The proposed filter outperforms alternative ones, leading to a 40% relative Word Error Rate (WER) reduction compared with the baseline Weighted Delay and Sum (WDAS) beamformer on the real test set, and a 15% relative WER reduction compared with the GEV-BAN method. The results also suggest that the speech recognition accuracy correlates more with the Mel-frequency cep-stral coefficients (MFCC) feature variance than with the noise reduction or the speech distortion level
Reference-based ranking procedure for environmental decision making: Insights from an ex-post analysis
No full textInternational audienc
Low-grazing angle propagation and scattering by an object above a highly-conducting rough sea surface in a ducting environment from an accelerated MoM
No full textInternational audienceIn a previous paper, by combining three techniques, i.e., Subdomain Decomposi-tion Iterative Method (SDIM), Adaptive Cross Approximation (ACA) and Forward-Backward Spectral Acceleration (FBSA), from the Method of Moments (MoM), ahigh-efficiency calculation of the propagation and scattering in ducting maritime en-vironments has been proposed. In this paper, this algorithm is updated by adding aperfectly-conducting object above the sea surface, assumed to be highly-conducting,which makes the environment very complex. Then, to quantify the effect of the objecton the total scattered field, the coherent and incoherent powers, with and withoutobject, are simulated by considering a surface of 800,000 unknowns (length of 6 kmand a frequency of 5 GHz)
Reduction of Qualitative Models of Biological Networks for Transient Dynamics Analysis
No full textInternational audienceQualitative models of dynamics of signalling pathways and gene regulatory networks allow to capture temporal properties of biological networks while requiring few parameters. However, these discrete models typically suffer from the so-called state space explosion problem which makes the formal assessment of their potential behaviours very challenging. In this paper, we describe a method to reduce a qualitative model for enhancing the tractability of analysis of transient reachability properties. The reduction does not change the dimension of the model, but instead limits its degree of freedom, therefore reducing the set of states and transitions to consider. We rely on a transition-centered specification of qualitative models by the mean of automata networks. Our framework encompass usual asynchronous Boolean and multi-valued network, as well as 1-bounded Petri nets. Applied to different large-scale biological networks from the litterature, we show that the reduction can lead to drastic improvement for the scalability of verification methods
A queueing model of an energy harvesting sensor node with data buffering
No full textInternational audienceBattery lifetime is a key impediment to long-lasting low power sensor nodes and networks thereof. Energy harvesting—conversion of ambient energy into electrical energy—has emerged as a viable alternative to battery power. Indeed, the harvested energy mitigates the dependency on battery power and can be used to transmit data. However, unfair data delivery delay and energy expenditure among sensors remain important issues for such networks. We study performance of sensor networks with mobile sinks: a mobile sink moves towards the transmission range of the different static sensor nodes to collect their data. We propose and analyse a Markovian queueing system to study the impact of uncertainty in energy harvesting, energy expenditure, data acquisition and data transmission. In particular, the energy harvesting sensor node is described by a system with two queues, one queue corresponding to the battery and the other to the data buffer. We illustrate our approach by numerical examples which show that energy harvesting correlation considerably affects performance measures like the mean data delay and the effective data collection rate
Adaptive RTO for Handshaking-based MAC Protocols in Underwater Acoustic Networks
No full textInternational audienceUnderwater acoustic networks (UANs) are attracting interest in recent decades.The unique characteristics of the underwater acoustic channel, such as long propagationdelay, delay variance, and high bit error rate, present challenges for themedium access control (MAC) protocol design in UANs. Most existing mediumaccess control protocols ignore the delay variance which prevents the accurateestimation of round trip time (RTT). The expected RTT value can be used to computethe Retransmission Time-Out (RTO) or the waiting time in MAC. The estimationof RTT is also meaningful for Automatic Repeat re-Quest (ARQ) schemebecause the system should ensure reliable data transmissions in the presence ofhigh bit error rate in the underwater acoustic channel. By analyzing the impact ofRTO on throughput under the effect of delay variance, we conclude that the fixedRTO is inefficient and RTO should be adaptively set to improve the throughput.We present a novel approach of predicting the RTT using a Bayesian dynamiclinear model, and then adjust RTO adaptively according to the predicted values.Simulation results show that the predicted values can adapt quickly to the sample RTT values. Under the effect of RTT fluctuations, the Bayesian algorithm offersperformance gains in terms of throughput and prediction performance, comparingwith Karn’s algorithm. Our study highlights the value of predicting the RTT usingBayesian approach in underwater acoustic networks
Criticality and propagation analysis of impacts between project deliverables
No full textInternational audienceThe implementation of a management by deliverablesand deadlines is based on detailed planning andstrict control of deliverables. It is a strategic decision thatreports to the project manager and a key element to thesuccess of complex projects. Based on the modeling ofthe project elements and their interactions using weighteddirected graphs, this article presents some contributionsto anticipate potential behavior of the project. Topologicaland propagation analyses are made to detect and prioritizecritical elements and critical interdependencies whileenlarging the sense of the polysemous word “critical.” Werecommend a set of topological indicators suitable for projectelements and interactions, which mainly allow us todiscuss “How the impact of a project element affects otherelements within the network? What is the collective influenceof this element?”. These indicators permit to prioritizeproject elements and their interactions by detectingthe most influential ones taking into account the networkstructure. For instance, they permit to evaluate the collectivecriticality of project deliverables and to re-evaluate thepriority of the risks associated with these deliverables bycoupling the traditional features of individual risks withthe topological indicators of the deliverables. Furthermore,some algorithms are applied to extract and visualize thepropagation path between two elements within the network.For example, this allows to provide a vision of potentialimpact propagation between two project deliverables, eitherthey are associated with two milestones or are critical. Anapplication to automotive industry illustrates the benefits ofthe approach, and some perspectives are drawn for furtherwork
Hidden Gaussian Markov Model for Distribued Fault Detection in Wireless Sensor Networks
No full textInternational audienceWireless Sensor Networks (WSN) are based on a large number of sensor nodesused to measure informations like temperature, acceleration, displacement orpressure. The measurements are used to estimate the state of the monitoredsystem or area. However, the quality of the measurements must be guaranteedto ensure the reliability of the estimated state of the system. Actually, sensorscan be used in a hostile environment such as, on a battle field in the presence offires, floods, earthquakes, for example. In these environments as well as in normaloperation,sensors can fail.The failure of sensor nodes can also be caused by other factors like: the failure ofmodule (such as sensing module) due to the fabrication process models, batterypower losts and so on. A WSN must be able to identify faulty nodes. Thereforewe propose a probabilistic approach based on Hidden Markov Model to identifyfaulty sensor nodes. Our proposed approach predicts the future state of each nodefrom its actual state, so the fault could be detect before it occurs. We use an aidedjudgment of neighbor sensor nodes in the network. The algorithm analyses thecorrelation of the sensors’ data with respect to its neighborhood. A systematicapproach to divide a network on cliques is proposed to fully draw the neighborhoodof each node in the network. After drawing the neighborhood of each node (cliques),damaged cliques are identified using Gaussian distribution theorem. Finally, we usethe Hidden Markov model to identify faulty nodes in the identified damaged cliquesby calculating the probability of each node to stay in its normal state. Simulationresults demonstrate our algorithm is efficient even for a huge wireless sensornetwork unlike previous approaches
Time Delay and Interface Roughness Estimation Using Modified ESPRIT With Interpolated Spatial Smoothing Technique
No full textInternational audienceIn civil engineering, ground penetrating radar is a common technique for evaluating the structure and quality of road pavement. This paper focuses on the estimation of the time delay and interface roughness of civil engineering structure, like pavements. The influence of interface roughness is taken into account in the signal model. A modified estimation of signal parameters via rotational invariance technique (ESPRIT) algorithm combined with an interpolated spatial smoothing technique is proposed. It allows us to jointly and efficiently estimate the time delay and interface roughness by ultrawideband radar (the upper frequency up to 8-10 GHz) with low computational complexity. The proposed algorithm is tested on both numerical and experimental data. Simulation and experimental results show the good performance of the proposed algorithm
Quantum confinement on non-complete Riemannian manifolds
No full text40 pages, 7 figures. (V2) corrected typos and updated referencesInternational audienceWe consider the quantum completeness problem, i.e. the problem of confining quantum particles, on a non-complete Riemannian manifold M equipped with a smooth measure ω, possibly degenerate or singular near the metric boundary of M , and in presence of a real-valued potential V ∈ L 2 loc (M). The main merit of this paper is the identification of an intrinsic quantity, the effective potential V eff , which allows to formulate simple criteria for quantum confinement. Let δ be the distance from the possibly non-compact metric boundary of M. A simplified version of the main result guarantees quantum completeness if V ≥ −cδ 2 far from the metric boundary and V eff + V ≥ 3 4δ 2 − κ δ , close to the metric boundary. These criteria allow us to: (i) obtain sharp quantum confinement results for measures with degeneracies or singularities near the metric boundary of M ; (ii) generalize the Kalf-Walter-Schmincke-Simon Theorem for strongly singular potentials to the Riemannian setting for any dimension of the singularity; (iii) give the first, to our knowledge, curvature-based criteria for self-adjointness of the Laplace-Beltrami operator; (iv) prove, under mild regularity assumptions, that the Laplace-Beltrami operator in almost-Riemannian geometry is essentially self-adjoint, partially settling a conjecture formulated in [9]