English

Compressed sensing and optimal denoising of monotone signals

Statistics Theory 2017-01-03 v1 Information Theory math.IT Statistics Theory

Abstract

We consider the problems of compressed sensing and optimal denoising for signals x0RN\mathbf{x_0}\in\mathbb{R}^N that are monotone, i.e., x0(i+1)x0(i)\mathbf{x_0}(i+1) \geq \mathbf{x_0}(i), and sparsely varying, i.e., x0(i+1)>x0(i)\mathbf{x_0}(i+1) > \mathbf{x_0}(i) only for a small number kk of indices ii. We approach the compressed sensing problem by minimizing the total variation norm restricted to the class of monotone signals subject to equality constraints obtained from a number of measurements Ax0A\mathbf{x_0}. For random Gaussian sensing matrices ARm×NA\in\mathbb{R}^{m\times N} we derive a closed form expression for the number of measurements mm required for successful reconstruction with high probability. We show that the probability undergoes a phase transition as mm varies, and depends not only on the number of change points, but also on their location. For denoising we regularize with the same norm and derive a formula for the optimal regularizer weight that depends only mildly on x0\mathbf{x_0}. We obtain our results using the statistical dimension tool.

Keywords

Cite

@article{arxiv.1701.00056,
  title  = {Compressed sensing and optimal denoising of monotone signals},
  author = {Eftychios A. Pnevmatikakis},
  journal= {arXiv preprint arXiv:1701.00056},
  year   = {2017}
}

Comments

To appear in the 42nd IEEE International Conference on Acoustics, Speech and Signal Processing ICASSP2017