Quantized Corrupted Sensing with Random Dithering

Abstract

Corrupted sensing concerns the problem of recovering a high-dimensional structured signal from a collection of measurements that are contaminated by unknown structured corruption and unstructured noise. In the case of linear measurements, the recovery performance of different convex programming procedures (e.g., generalized Lasso and its variants) is well established in the literature. However, in practical applications of digital signal processing, the quantization process is inevitable, which often leads to non-linear measurements. This paper is devoted to studying corrupted sensing under quantized measurements. Specifically, we demonstrate that, with the aid of uniform dithering, both constrained and unconstrained Lassos are able to recover signal and corruption from the quantized samples when the measurement matrix is sub-Gaussian. Our theoretical results reveal the role of quantization resolution in the recovery performance of Lassos. Numerical experiments are provided to confirm our theoretical results.

arXiv comment13 pages, 19 figures

References in corpus (9)

• The Convex Geometry of Linear Inverse Problems
• Rank-Sparsity Incoherence for Matrix Decomposition
• The LASSO with Non-linear Measurements is Equivalent to One With Linear Measurements
• Corrupted Sensing: Novel Guarantees for Separating Structured Signals
• Robust and Low-Rank Representation for Fast Face Identification with Occlusions
• Convexity in source separation: Models, geometry, and algorithms
• Low-Rank Positive Semidefinite Matrix Recovery from Corrupted Rank-One Measurements
• Uncertainty Relations and Sparse Signal Recovery for Pairs of General Signal Sets
• Uniform Recovery Bounds for Structured Random Matrices in Corrupted Compressed Sensing