Magneto- and electroencephalography (M/EEG) measure the electromagnetic fields produced by neural electrical currents. Given a conductor model for head, distribution of source currents in brain, Maxwell's equations allow one to compute ensuing M/EEG signals. actual measurements solution this forward problem, can localize, space time, brain regions that have recorded data. However, due physics limited number sensors compared possible locations, measurement noise, inverse problem is ill-posed....

Sparse and structured signal expansions on dictionaries can be obtained through explicit modeling in the coefficient domain. The originality of present article lies construction study generalized shrinkage operators, whose goal is to identify significance maps give rise thresholding. These generalize Group-Lasso previously introduced Elitist Lasso by introducing more flexibility domain modeling, lead notion social sparsity. proposed operators are studied theoretically embedded iterative...

We consider the audio declipping problem by using iterative thresholding algorithms and principle of social sparsity. This recently introduced approach features thresholding/shrinkage operators which allow to model dependencies between neighboring coefficients in expansions with time-frequency dictionaries. A new unconstrained convex formulation is introduced. The chosen structured are so called windowed group-Lasso persistent empirical Wiener. usage these significantly improves quality...

We consider the problem of extracting source signals from an under-determined convolutive mixture assuming known mixing filters. State-of-the-art methods operate in time-frequency domain and rely on narrowband approximation process by complex-valued multiplication each frequency bin. The are then estimated minimizing either a fitting cost or <i xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">ℓ</i> <sub...

To display the time and frequency content of a given signal large variety techniques exist.In this paper, we give an overview linear time-frequency representations, focusing mainly on two fundamental aspects.The first one is introduction flexibility, more precisely construction waveform systems that can be adapted to specific signals, or processing problems.To do this, base constructions frame theory, which allows lot options, while still ensuring perfect reconstruction.The second aspect...

We consider the problem of blind source separation for underdetermined convolutive mixtures. Based on multiplicative narrowband approximation in time-frequency domain with help short-time-Fourier-transform (STFT) and sparse representation signals, we formulate an optimization framework. This framework is then generalized based recently investigated statistics room impulse response. Algorithms convergence proof are employed to solve proposed problems. The evaluation frameworks algorithms...

We address the problem of learning classifiers using several kernel functions. On contrary to many contributions in field from different sources information kernels, we here do not assume that kernels used are positive definite. The interested involves a misclassification loss term and regularization is expressed by means mixed norm. use norm allows us enforce some sparsity structure, particular case which is, for instance, Group Lasso. solve convex employing proximal minimization...

Imaging inverse problems can be formulated as an optimization problem and solved thanks to algorithms such forward-backward or ISTA (Iterative Shrinkage/Thresholding Algorithm) for which non smooth functionals with sparsity constraints minimized efficiently. However, the soft thresholding operator involved in this algorithm leads a biased estimation of large coefficients. That is why step allowing reduce bias introduced practice. Indeed, statistical community, variety operators have been...

The inverse problem with distributed dipoles models in M/EEG is strongly ill-posed requiring to set priors on the solution. Most common are based a convenient lscr <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sub> norm. However such methods known smear estimated distribution of cortical currents. In order provide sparser solutions, other norms than have been proposed literature, but they often do not pass test real data. Here we propose...

In this paper, we propose a new unconstrained nonnegative matrix factorization method designed to utilize the multilayer structure of audio signals improve quality source separation. The tonal layer is sparse in frequency and temporally stable, while transient composed short term broadband sounds. Our has part well suited for extraction which decomposes orthogonal components, represented by regular decomposition. Experiments on synthetic real music data separation context show that such...

A new approach for signal expansion with respect to hybrid dictionaries, based upon probabilistic modeling is proposed and studied. The modeled as a sparse linear combination of waveforms, taken from the union two orthonormal bases, random coefficients. behavior analysis coefficients, namely inner products all basis functions, studied in details, which shows that these coefficients may generally be classified categories: significant versus insignificant Conditions ensuring feasibility such...

In this paper, we propose a supervised multilayer factorization method designed for harmonic/percussive source separation and drum extraction. Our decomposes the audio signals in sparse orthogonal components which capture harmonic content, while is represented by an extension of non negative matrix able to exploit time-frequency dictionaries take into account stationary sounds. The represent various real hits decomposition has more physical sense allows better interpretation results....

One of the most general models music signals considers that such can be represented as a sum two distinct components: tonal part is sparse in frequency and temporally stable transient (or percussive) composed short-term broadband sounds. In this paper, we propose novel hybrid method built upon nonnegative matrix factorization (NMF) decomposes time representation an audio signal into components. The estimated by orthogonal decomposition, straightforward NMF decomposition constrained...

We consider the problem of extracting source signals from an under-determined convolutive mixture, assuming known filters. start its formulation as a minimization convex functional, combining classical I <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sub> discrepancy term between observed mixture and one reconstructed estimated sources, sparse regularization coefficients in time-frequency domain. then introduce first kind structure, using...

Independent component analysis (ICA) has been a major tool for blind source separation (BSS). Both theoretical and practical evaluations showed that the hypothesis of independence suits well audio signals. In last few years, optimization approach based on sparsity emerged as another efficient implement BSS. This paper starts from introducing some new BSS methods take advantages both decorrelation (which is direct consequence independence) using overcomplete Gabor representation. It shown...