Spikes super-resolution with random Fourier sampling

We leverage recent results from machine learning to show theoretically and practically that it is possible to stably recover a signal made of few spikes (in the gridless setting) from few random weighted Fourier measurements. Given a free choice of frequencies, a number of measurements lower than with the traditional low-pass filter (uniform sampling of low frequencies) guarantees stable recovery.

[1]  Patrick Pérez,et al.  Sketching for Large-Scale Learning of Mixture Models. (Apprentissage de modèles de mélange à large échelle par Sketching) , 2017 .

[2]  Patrick Pérez,et al.  Fundamental Performance Limits for Ideal Decoders in High-Dimensional Linear Inverse Problems , 2013, IEEE Transactions on Information Theory.

[3]  R. DeVore,et al.  A Simple Proof of the Restricted Isometry Property for Random Matrices , 2008 .

[4]  Thierry Blu,et al.  Towards Generalized FRI Sampling With an Application to Source Resolution in Radioastronomy , 2017, IEEE Transactions on Signal Processing.

[5]  Rémi Gribonval,et al.  Compressive K-means , 2017, 2017 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP).

[6]  Didier Henrion,et al.  Exact Solutions to Super Resolution on Semi-Algebraic Domains in Higher Dimensions , 2015, IEEE Transactions on Information Theory.

[7]  Gabriel Peyré,et al.  Exact Support Recovery for Sparse Spikes Deconvolution , 2013, Foundations of Computational Mathematics.

[8]  Gongguo Tang,et al.  Resolution limits for atomic decompositions via Markov-Bernstein type inequalities , 2015, 2015 International Conference on Sampling Theory and Applications (SampTA).

[9]  Rémi Gribonval,et al.  Recipes for stable linear embeddings from Hilbert spaces to R^m , 2015 .

[10]  Emmanuel J. Candès,et al.  Super-Resolution from Noisy Data , 2012, Journal of Fourier Analysis and Applications.