Projection Design For Statistical Compressive Sensing: A Tight Frame Based Approach

TL;DR

This paper introduces a novel framework for designing sensing matrices in compressive sensing that leverage tight frames to optimize mean squared error performance under sensing cost constraints, providing closed-form solutions with superior empirical results.

Contribution

It proposes a new tight frame-based sensing matrix design method that is non-iterative and outperforms existing designs in various sparse recovery scenarios.

Findings

Designs are closed form and computationally efficient.

Proposed matrices outperform existing designs in numerical tests.

Effective across multiple sparse recovery algorithms.

Abstract

In this paper, we develop a framework to design sensing matrices for compressive sensing applications that lead to good mean squared error (MSE) performance subject to sensing cost constraints. By capitalizing on the MSE of the oracle estimator, whose performance has been shown to act as a benchmark to the performance of standard sparse recovery algorithms, we use the fact that a Parseval tight frame is the closest design - in the Frobenius norm sense - to the solution of a convex relaxation of the optimization problem that relates to the minimization of the MSE of the oracle estimator with respect to the equivalent sensing matrix, subject to sensing energy constraints. Based on this result, we then propose two sensing matrix designs that exhibit two key properties: i) the designs are closed form rather than iterative; ii) the designs exhibit superior performance in relation to other designs in the literature, which is revealed by our numerical investigation in various scenarios with different sparse recovery algorithms including basis pursuit de-noise (BPDN), the Dantzig selector and orthogonal matching pursuit (OMP).