Matthieu Kowalski

Mixed-norm estimates for the M/EEG inverse problem using accelerated gradient methods

Magneto- and electroencephalography (M/EEG) measure the electromagnetic fields produced by the neural electrical currents. Given a conductor model for the head, and the distribution of source currents in the brain, Maxwells equations allow one to compute the ensuing M/EEG signals. Given the actual M/EEG measurements and the solution of this forward problem, one can localize, in space and in time, the brain regions that have produced the recorded data. However, due to the physics of the problem, the limited number of sensors compared to the number of possible source locations, and measurement noise, this inverse problem is ill-posed. Consequently, additional constraints are needed. Classical inverse solvers, often called minimum norm estimates (MNE), promote source estimates with a small ℓ₂ norm. Here, we consider a more general class of priors based on mixed norms. Such norms have the ability to structure the prior in order to incorporate some additional assumptions about the sources. We refer to such solvers as mixed-norm estimates (MxNE). In the context of M/EEG, MxNE can promote spatially focal sources with smooth temporal estimates with a two-level ℓ₁/ℓ₂ mixed-norm, while a three-level mixed-norm can be used to promote spatially non-overlapping sources between different experimental conditions. In order to efficiently solve the optimization problems of MxNE, we introduce fast first-order iterative schemes that for the ℓ₁/ℓ₂ norm give solutions in a few seconds making such a prior as convenient as the simple MNE. Furthermore, thanks to the convexity of the optimization problem, we can provide optimality conditions that guarantee global convergence. The utility of the methods is demonstrated both with simulations and experimental MEG data.

Time-Frequency Mixed-Norm Estimates: Sparse M/EEG imaging with non-stationary source activations

Magnetoencephalography (MEG) and electroencephalography (EEG) allow functional brain imaging with high temporal resolution. While solving the inverse problem independently at every time point can give an image of the active brain at every millisecond, such a procedure does not capitalize on the temporal dynamics of the signal. Linear inverse methods (minimum-norm, dSPM, sLORETA, beamformers) typically assume that the signal is stationary: regularization parameter and data covariance are independent of time and the time varying signal-to-noise ratio (SNR). Other recently proposed non-linear inverse solvers promoting focal activations estimate the sources in both space and time while also assuming stationary sources during a time interval. However such a hypothesis holds only for short time intervals. To overcome this limitation, we propose time-frequency mixed-norm estimates (TF-MxNE), which use time-frequency analysis to regularize the ill-posed inverse problem. This method makes use of structured sparse priors defined in the time-frequency domain, offering more accurate estimates by capturing the non-stationary and transient nature of brain signals. State-of-the-art convex optimization procedures based on proximal operators are employed, allowing the derivation of a fast estimation algorithm. The accuracy of the TF-MxNE is compared with recently proposed inverse solvers with help of simulations and by analyzing publicly available MEG datasets.

Social Sparsity! Neighborhood Systems Enrich Structured Shrinkage Operators

Sparse and structured signal expansions on dictionaries can be obtained through explicit modeling in the coefficient domain. The originality of the present article lies in the construction and the study of generalized shrinkage operators, whose goal is to identify structured significance maps and give rise to structured thresholding. These generalize Group-Lasso and the previously introduced Elitist Lasso by introducing more flexibility in the coefficient domain modeling, and lead to the notion of social sparsity. The proposed operators are studied theoretically and embedded in iterative thresholding algorithms. Moreover, a link between these operators and a convex functional is established. Numerical studies on both simulated and real signals confirm the benefits of such an approach.

Beyond the Narrowband Approximation: Wideband Convex Methods for Under-Determined Reverberant Audio Source Separation

We consider the problem of extracting the source signals from an under-determined convolutive mixture assuming known mixing filters. State-of-the-art methods operate in the time-frequency domain and rely on narrowband approximation of the convolutive mixing process by complex-valued multiplication in each frequency bin. The source signals are then estimated by minimizing either a mixture fitting cost or a ℓ1 source sparsity cost, under possible constraints on the number of active sources. In this paper, we define a wideband ℓ2 mixture fitting cost circumventing the above approximation and investigate the use of a ℓ1,2 mixed-norm cost promoting disjointness of the source time-frequency representations. We design a family of convex functionals combining these costs and derive suitable optimization algorithms. Experiments indicate that the proposed wideband methods result in a signal-to-distortion ratio improvement of 2 to 5 dB compared to the state-of-the-art on reverberant speech mixtures.

Functional brain imaging with M/EEG using structured sparsity in time-frequency dictionaries

Magnetoencephalography (MEG) and electroencephalography (EEG) allow functional brain imaging with high temporal resolution. While time-frequency analysis is often used in the field, it is not commonly employed in the context of the ill-posed inverse problem that maps the MEG and EEG measurements to the source space in the brain. In this work, we detail how convex structured sparsity can be exploited to achieve a principled and more accurate functional imaging approach. Importantly, time-frequency dictionaries can capture the non-stationary nature of brain signals and state-of-the-art convex optimization procedures based on proximal operators allow the derivation of a fast estimation algorithm. We compare the accuracy of our new method to recently proposed inverse solvers with help of simulations and analysis of real MEG data.

Adapted and Adaptive Linear Time-Frequency Representations: A Synthesis Point of View

A large variety of techniques exist to display the time and frequency content of a given signal. In this article, we give an overview of linear time-frequency representations, focusing mainly on two fundamental aspects. The first is the introduction of flexibility, more precisely, the construction of time-frequency waveform systems that can be adapted to specific signals or specific signal processing problems. To do this, we base the constructions on frame theory, which allows many options while still ensuring perfect reconstruction. The second aspect is the choice of the synthesis framework rather than the usual analysis framework. Instead of considering the correlation, i.e. the inner product, of the signal with the chosen waveforms, we find appropriate coefficients in a linear combination of those waveforms to synthesize the given signal. We show how this point of view allows the easy introduction of prior information into the representation. We give an overview of methods for transform domain modeling, in particular those based on sparsity and structured sparsity. Finally, we present an illustrative application for these concepts: a denoising scheme.

Audio declipping with social sparsity

We consider the audio declipping problem by using iterative thresholding algorithms and the 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 of the audio declipping problem is introduced. The chosen structured thresholding operators are the so called windowed group-Lasso and the persistent empirical Wiener. The usage of these operators significantly improves the quality of the reconstruction, compared to simple soft-thresholding. The resulting algorithm is fast, simple to implement, and it outperforms the state of the art in terms of signal to noise ratio.

Random Models for Sparse Signals Expansion on Unions of Bases With Application to Audio Signals

A new approach for signal expansion with respect to hybrid dictionaries, based upon probabilistic modeling is proposed and studied. The signal is modeled as a sparse linear combination of waveforms, taken from the union of two orthonormal bases, with random coefficients. The behavior of the analysis coefficients, namely inner products of the signal with all basis functions, is studied in details, which shows that these coefficients may generally be classified in two categories: significant coefficients versus insignificant coefficients. Conditions ensuring the feasibility of such a classification are given. When the classification is possible, it leads to efficient estimation algorithms, that may in turn be used for denoising or coding purposes. The proposed approach is illustrated by numerical experiments on audio signals, using MDCT bases. However, it is general enough to be applied without much modifications in different contexts, for example in image processing.

Thresholding RULES and iterative shrinkage/thresholding algorithm: A convergence study

Imaging inverse problems can be formulated as an optimization problem and solved thanks to algorithms such as forward-backward or ISTA (Iterative Shrinkage/Thresholding Algorithm) for which non smooth functionals with sparsity constraints can be minimized efficiently. However, the soft thresholding operator involved in this algorithm leads to a biased estimation of large coefficients. That is why a step allowing to reduce this bias is introduced in practice. Indeed, in the statistical community, a large variety of thresholding operators have been studied to avoid the biased estimation of large coefficients; for instance, the non negative Garrote or the the SCAD thresholding. One can associate a non convex penalty to these operators. We study the convergence properties of ISTA, possibly relaxed, with any thresholding rule and show that they correspond to a semi-convex penalty. The effectiveness of this approach is illustrated on image inverse problems.

An unified approach for blind source separation using sparsity and decorrelation

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 for audio signals. In the last few years, optimization approach based on sparsity has emerged as another efficient implement for BSS. This paper starts from introducing some new BSS methods that take advantages of both decorrelation (which is a direct consequence of independence) and sparsity using overcomplete Gabor representation. It is shown that the proposed methods work in both under-determined and over-determined cases. Experimental results illustrate the good performances of these approaches for audio mixtures.