Publications

2017

• Supervised Group Nonnegative Matrix Factorisation With Similarity Constraints And Applications To Speaker Identification
  
  • Serizel Romain
  • Bisot Victor
  • Essid Slim
  • Richard Gael
  , 2017. This paper presents supervised feature learning approaches for speaker identification that rely on nonnegative matrix factorisa-tion. Recent studies have shown that group nonnegative matrix factorisation and task-driven supervised dictionary learning can help performing effective feature learning for audio classification problems. This paper proposes to integrate a recent method that relies on group nonnegative matrix factorisation into a task-driven supervised framework for speaker identification. The goal is to capture both the speaker variability and the session variability while exploiting the discriminative learning aspect of the task-driven approach. Results on a subset of the ESTER corpus prove that the proposed approach can be competitive with I-vectors. Index Terms— Nonnegative matrix factorisation, feature learning , dictionary learning, online learning, speaker identification
• Parallelized Stochastic Gradient Markov Chain Monte Carlo Algorithms for Non-Negative Matrix Factorization
  
  • Şimşekli Umut
  • Durmus Alain
  • Badeau Roland
  • Richard Gael
  • Moulines Éric
  • Cemgil Taylan
  , 2017. Stochastic Gradient Markov Chain Monte Carlo (SG-MCMC) methods have become popular in modern data analysis problems due to their computational efficiency. Even though they have proved useful for many statistical models, the application of SG-MCMC to non- negative matrix factorization (NMF) models has not yet been extensively explored. In this study, we develop two parallel SG-MCMC algorithms for a broad range of NMF models. We exploit the conditional independence structure of the NMF models and utilize a stratified sub-sampling approach for enabling parallelization. We illustrate the proposed algorithms on an image restoration task and report encouraging results.
• Multichannel audio source separation: variational inference of time-frequency sources from time-domain observations
  
  • Leglaive Simon
  • Badeau Roland
  • Richard Gael
  , 2017. A great number of methods for multichannel audio source separation are based on probabilistic approaches in which the sources are modeled as latent random variables in a time-frequency (TF) domain. For reverberant mixtures, most of the methods approximate the time-domain convolutive mixing process in the TF-domain, assuming short mixing filters. The TF latent sources are then inferred from the TF mixture observations. In this paper we propose to infer latent TF sources from the time-domain observations. This approach allows us to exactly model the convolutive mixing process. The inference procedure rely on a variational expectation-maximization algorithm. In significant reverberation conditions, we show that our approach leads a Signal-to-Distortion Ratio improvement of 5.5 dB.
• Motion informed audio source separation
  
  • Parekh Sanjeel
  • Essid Slim
  • Ozerov Alexey
  • Duong Ngoc
  • Pérez Patrick
  • Richard Gael
  , 2017. In this paper we tackle the problem of single channel audio source separation driven by descriptors of the sounding object's motion. As opposed to previous approaches, motion is included as a soft-coupling constraint within the nonnegative matrix factorization framework. The proposed method is applied to a multimodal dataset of instruments in string quartet performance recordings where bow motion information is used for separation of string instruments. We show that the approach offers better source separation result than an audio-based baseline and the state-of-the-art multimodal-based approaches on these very challenging music mixtures.
• A Railroad Detection Algorithm for Infrastructure Surveillance using Enduring Airborne Systems
  
  • Purica Andrei
  • Pesquet-Popescu Beatrice
  • Dufaux Frederic
  , 2017. Infrastructure surveillance is an important requirement for many companies. With the advancement of technology, drones can now provide an efficient tool for such applications. A possible future scenario is the automated surveillance of railroads. Whereas numerous algorithms that provide railroad detection exist, they have mainly focused either on satellite images or for small, low altitude drones which are unsuitable for our particular scenario. In this paper we propose a railroad detection algorithm tailored for large, high altitude enduring drones. More specifically, we use Hough Transform to detect lines and perform a line clustering in the Rho and Theta space. A score model is also proposed in order to identify the railroad. We test our method on several sequences supplied by Airbus Defense & Space and show our algorithm to provide a detection rate of 93.23% in average. (10.1109/icassp.2017.7952544)
  DOI : 10.1109/icassp.2017.7952544
• Overlapping sound event detection with supervised Nonnegative Matrix Factorization
  
  • Bisot Victor
  • Essid Slim
  • Richard Gael
  , 2017, pp.31-35. (10.1109/ICASSP.2017.7951792)
  DOI : 10.1109/ICASSP.2017.7951792
• Drum extraction in single channel audio signals using multi-layer non negative matrix factor deconvolution
  
  • Laroche Clément
  • Papadopoulos Hélène
  • Kowalski Matthieu
  • Richard Gael
  , 2017. In this paper, we propose a supervised multilayer factorization method designed for harmonic/percussive source separation and drum extraction. Our method decomposes the audio signals in sparse orthogonal components which capture the harmonic content, while the drum is represented by an extension of non negative matrix factorization which is able to exploit time-frequency dictionaries to take into account non stationary drum sounds. The drum dictionaries represent various real drum hits and the decomposition has more physical sense and allows for a better interpretation of the results. Experiments on real music data for a harmonic/percussive source separation task show that our method outperforms other state of the art algorithms. Finally, our method is very robust to non stationary harmonic sources that are usually poorly decomposed by existing methods. (10.1109/icassp.2017.7952115)
  DOI : 10.1109/icassp.2017.7952115
• Good Features to Track for RGBD images
  
  • Karpushin Maxim
  • Valenzise Giuseppe
  • Dufaux Frederic
  , 2017. RGBD (texture-plus-depth) image representation enriches traditional 2D content with additional geometrical information, having the potential to improve the performance of many computer vision tasks. In image matching, this has been partially studied by considering how depth maps can help render feature descriptors more distinctive. However, little has been done to design keypoint detection approaches able to leverage the availability of depth information. In this paper, we propose a novel and robust approach for detecting corners from RGBD images. Our method modifies a classical corner detection strategy, based on local second-order moment matrices, by computing derivatives in a coordinate system which reflects the local properties of object surfaces. Our results demonstrate a higher stability to out-of-plane rotations of the proposed RGBD corner detector both in terms of feature repeatability and in a visual odometry application. (10.1109/icassp.2017.7952473)
  DOI : 10.1109/icassp.2017.7952473
• Parametric estimation of spectrum driven by an exogenous signal
  
  • Dupré La Tour Tom
  • Grenier Yves
  • Gramfort Alexandre
  , 2017. In this paper, we introduce new parametric generative driven auto-regressive (DAR) models. DAR models provide a non-linear and non-stationary spectral estimation of a signal, conditionally to another exogenous signal. We detail how inference can be done efficiently while guaranteeing model stability. We show how model comparison and hyper-parameter selection can be done using likelihood estimates. We also point out the limits of DAR models when the exogenous signal contains too high frequencies. Finally, we illustrate how DAR models can be applied on neuro-physiologic signals to characterize phase-amplitude coupling.
• Alpha-Stable Multichannel Audio Source Separation
  
  • Leglaive Simon
  • Şimşekli Umut
  • Liutkus Antoine
  • Badeau Roland
  • Richard Gael
  , 2017. In this paper, we focus on modeling multichannel audio signals in the short-time Fourier transform domain for the purpose of source separation. We propose a probabilistic model based on a class of heavy-tailed distributions, in which the observed mixtures and the latent sources are jointly modeled by using a certain class of multivariate alpha-stable distributions. As opposed to the conventional Gaussian models, where the observations are constrained to lie just within a few standard deviations near the mean, the pro- posed heavy-tailed model allows us to account for spurious data or important uncertainties in the model. We develop a Monte Carlo Expectation-Maximization algorithm for making inference in the proposed model. We show that our approach leads to significant improvements in audio source separation under corrupted mixtures and in spatial audio object coding.
• Phase-dependent anisotropic Gaussian model for audio source separation
  
  • Magron Paul
  • Badeau Roland
  • David Bertrand
  , 2017. Phase reconstruction of complex components in the time-frequency domain is a challenging but necessary task for audio source separation. While traditional approaches do not exploit phase constraints that originate from signal modeling, some prior information about the phase can be obtained from sinusoidal modeling. In this paper, we introduce a probabilistic mixture model which allows us to incorporate such phase priors within a source separation framework. While the magnitudes are estimated beforehand, the phases are modeled by Von Mises random variables whose location parameters are the phase priors. We then approximate this non-tractable model by an anisotropic Gaussian model, in which the phase dependencies are preserved. This enables us to derive an MMSE estimator of the sources which optimally combines Wiener filtering and prior phase estimates. Experimental results highlight the potential of incorporating phase priors into mixture models for separating overlapping components in complex audio mixtures.
• A Non-Monte-Carlo Methodology for Variability Analysis of Magnetic Tunnel Junction-Based Circuits
  
  • Wang You
  • Cai Hao
  • Naviner Lirida
  • Zhao Weisheng
  IEEE Transactions on Magnetics, Institute of Electrical and Electronics Engineers, 2017, 53 (3), pp.1-6. (10.1109/TMAG.2016.2638913)
  DOI : 10.1109/TMAG.2016.2638913
• Active Phased Antenna Arrays Calibration Method Including Edge Effects and Mutual Coupling
  
  • Chalumyan Taguhi
  • Maas Olivier
  • Begaud Xavier
  • Huyart Bernard
  , 2017.
• François Lefeuvre: A long Way among Waves
  
  • Tanzi Tullio
  • Hamelin Joel
  Radio Science Bulletin, Union Radio-Scientifique Internationale (URSI), 2017 (360), pp.69-74.
• Statistical Modeling of the Reflection Coefficient of deformable antennas
  
  • Du Jinxin
  • Roblin Christophe
  , 2017. A modeling methodology is proposed for characterizing the reflection coefficient S11( f ) of narrow band antennas undergoing random disturbances. Firstly, identification techniques are used to get a parsimonious representation of the S11; then the Polynomial Chaos Expansion (PCE) method is used to characterize quantitatively the influence of random disturbances on the compressed S11. The derived S11 model can be used as efficient surrogate for statistical analysis of antennas’ frequency behavior. We have applied the proposed methodology to two narrow band antennas – a deformable dipole and a textile patch – in order to demonstrate its performance. Models with good accuracy have been derived for both cases.
• Software Stack for Code Generation of Software-Defined Radio
  
  • Canuel Adrien
  • Pacalet Renaud
  • Apvrille Ludovic
  , 2017. Hardware architecture for signal processing have an increasing heterogeneity and parallelism in order to cope with the real time performance and low consumption problematics. This complexity has an influence on the development of application running on those platforms. To ease the development processes, modeling tools already exists. Our laboratory created its own modeling tool, which can generate a simulation code from the mode l designed. The work presented in this article describe an abstraction methodology of the hardware for this tool, so that the code generated by this tool can run directly on the target platform. This methodology relies on the definition of data structures and functions directly included in the code by the generator, as well as services that must be provided by the runtime environment.
• URSI France 2017 Workshop Radio Science for Humanity
  
  • Tanzi Tullio
  Radio Science Bulletin, Union Radio-Scientifique Internationale (URSI), 2017 (360), pp.62-68.
• Learning Attribute Representations for Remote Sensing Ship Category Classification
  
  • Oliveau Quentin
  • Sahbi Hichem
  IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, IEEE, 2017, 10 (6), pp.2830-2840.
• Dichotomic Sphere Decoder
  
  • Khsiba Mohamed-Achraf
  • Rekaya-Ben Othman Ghaya
  , 2017.
• AVC to HEVC transcoder based on quadtree limitation
  
  • Mora Elie Gabriel
  • Cagnazzo Marco
  • Dufaux Frederic
  Multimedia Tools and Applications, Springer Verlag, 2017, 76 (6), pp.8991-9015. Following the finalization of the state-of-the-art High Efficiency Video Coding (HEVC) standard in January 2013, several new services are being deployed in order to take advantage of the superior coding efficiency (estimated at 50 % less bitrate for the same visual quality) that this standard provides over its predecessor: H.264 / Advanced Video Coding (AVC). However, the switch from AVC to HEVC is not trivial as most video content is still encoded in AVC. Consequently, there is a growing need for fast AVC to HEVC transcoders in the market today. While a trivial transcoder can be made by simply cascading an AVC decoder and an HEVC encoder, fast transcoding cannot be achieved. In this paper, we present an AVC to HEVC transcoder where decoded AVC blocks are first fused according to their motion similarity. The resulting fusion map is then used to limit the quadtree of HEVC coded frames. AVC motion vectors are also used to determine a better starting point for integer motion estimation. Experimental results show that significant transcoder execution time savings of 63 % can be obtained with only a 1.4 % bitrate increase compared to the trivial transcoder. (10.1007/s11042-016-3498-8)
  DOI : 10.1007/s11042-016-3498-8
• Cybersecurity and Privacy Solutions in Smart Cities
  
  • Khatoun Rida
  • Zeadally Sherali
  IEEE Communications Magazine, Institute of Electrical and Electronics Engineers, 2017, 55 (3), pp.51-59. (10.1109/MCOM.2017.1600297CM)
  DOI : 10.1109/MCOM.2017.1600297CM
• AVC to HEVC transcoder based on quadtree limitation
  
  • Mora Elie Gabriel
  • Cagnazzo Marco
  • Dufaux Frederic
  Multimedia Tools and Applications, Springer Verlag, 2017, 76 (6), pp.8991-9015. Following the finalization of the state-of-the-art High Efficiency Video Coding (HEVC) standard in January 2013, several new services are being deployed in order to take advantage of the superior coding efficiency (estimated at 50 % less bitrate for the same visual quality) that this standard provides over its predecessor: H.264 / Advanced Video Coding (AVC). However, the switch from AVC to HEVC is not trivial as most video content is still encoded in AVC. Consequently, there is a growing need for fast AVC to HEVC transcoders in the market today. While a trivial transcoder can be made by simply cascading an AVC decoder and an HEVC encoder, fast transcoding cannot be achieved. In this paper, we present an AVC to HEVC transcoder where decoded AVC blocks are first fused according to their motion similarity. The resulting fusion map is then used to limit the quadtree of HEVC coded frames. AVC motion vectors are also used to determine a better starting point for integer motion estimation. Experimental results show that significant transcoder execution time savings of 63 % can be obtained with only a 1.4 % bitrate increase compared to the trivial transcoder.
• Controllable Rare Events in an Optically-Injected Fabry-Perot Semiconductor Laser
  
  • Schires Kevin
  • Grillot Frédéric
  , 2017.
• Conception, réalisation et caractérisation d’absorbants électromagnétiques à métamatériaux pour applications spatiales et aéronautiques
  
  • Begaud Xavier
  • Lepage A. C.
  , 2017.
• Robust Adaptive Detection of Buried Pipes using GPR
  
  • Hoarau Q
  • Ginolhac G
  • Atto A M
  • Nicolas Jean-Marie
  Signal Processing, Elsevier, 2017, 132, pp.293–305. Detection of buried objects such as pipes using a Ground Penetrating Radar (GPR) is intricate for three main reasons. First, noise is important in the resulting image because of the presence of several rocks and/or layers in the ground, highly influencing the Probability of False Alarm (PFA) level. Also, wave speed and object responses are unknown in the ground and depend on the relative permit-tivity, which is not directly measurable. Finally, the depth of the pipes leads to strong attenuation of the echoed signal, leading to poor SNR scenarios. In this paper, we propose a detection method: (1) enhancing the signal of interest while reducing the noise and layer contributions, and (2) giving a local estimate of the relative permittivity. We derive an adaptive detector where the signal of interest is parametrised by the wave speed in the ground. For this detector, noise is assumed to follow a Spherically Invariant Random Vector (SIRV) distribution in order to obtain a robust detection. We use robust maximum likelihood-type covariance matrix estimators called M-estimators. To handle the significant amount of data, we consider regularised versions of said estimators. Simulation will allow to estimate the relation PFA-Threshold. Comparison is performed with standard GPR processing methods, showing the aptitude of the method in detecting pipes having low response levels with a reasonable PFA. (10.1016/j.sigpro.2016.07.001)
  DOI : 10.1016/j.sigpro.2016.07.001