Abstract:We consider the problem of recovering two unknown vectors, $\boldsymbol{w}$ and $\boldsymbol{x}$, of length $L$ from their circular convolution. We make the structural assumption that the two vectors are members of known subspaces, one with dimension $N$ and the other with dimension $K$. Although the observed convolution is nonlinear in both $\boldsymbol{w}$ and $\boldsymbol{x}$, it is linear in the rank-1 matrix formed by their outer product $\boldsymbol{w}\boldsymbol{x}^*$. This observation allows us to recast the deconvolution problem as low-rank matrix recovery problem from linear measurements, whose natural convex relaxation is a nuclear norm minimization program. We prove the effectiveness of this relaxation by showing that for "generic" signals, the program can deconvolve $\boldsymbol{w}$ and $\boldsymbol{x}$ exactly when the maximum of $N$ and $K$ is almost on the order of $L$. That is, we show that if $\boldsymbol{x}$ is drawn from a random subspace of dimension $N$, and $\boldsymbol{w}$ is a vector in a subspace of dimension $K$ whose basis vectors are "spread out" in the frequency domain, then nuclear norm minimization recovers $\boldsymbol{w}\boldsymbol{x}^*$ without error. We discuss this result in the context of blind channel estimation in communications. If we have a message of length $N$ which we code using a random $L\times N$ coding matrix, and the encoded message travels through an unknown linear time-invariant channel of maximum length $K$, then the receiver can recover both the channel response and the message when $L\gtrsim N+K$, to within constant and log factors.

Deconvolution with Unknown Error Distribution Interpreted as Blind Isotonic Regression

Multi-frame Blind Deconvolution Using Sparse Priors

Identifiability of Multichannel Blind Deconvolution and Nonconvex Regularization Algorithm

Blind Deconvolution Using an Improved L0 Sparse Representation

The Effects of Error Magnitude and Bandwidth Selection for Deconvolution with Unknown Error Distribution.

Blind Deconvolution on Graphs: Exact and Stable Recovery

Blind Deconvolution using Convex Programming

A Novel Approach Based on Differential Evolution for Blind Deconvolution.

Deconvolution of repeated measurements corrupted by unknown noise

Rapid, Robust, and Reliable Blind Deconvolution Via Nonconvex Optimization

Fundamental Limits of Blind Deconvolution Part I: Ambiguity Kernel

Blind Deconvolution of Sparse Graph Signals in the Presence of Perturbations

Support and distribution inference from noisy data

Single Image Blind Deconvolution Using Sparse Representation and Structural Self-similarity

Multichannel Sparse Blind Deconvolution on the Sphere

Blind Image Deblurring Using Spectral Properties of Convolution Operators

Robust Statistics for Image Deconvolution

How robust is randomized blind deconvolution via nuclear norm minimization against adversarial noise?

Linear regression with unmatched data: a deconvolution perspective

Multiplicative deconvolution under unknown error distribution

Kernel Diffusion: An Alternate Approach to Blind Deconvolution