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.

Blind Deconvolution in Nonminimum Phase Systems Using Cascade Structure

Multi-frame Blind Deconvolution Using Sparse Priors

Lp (1≤p≤∞) BLIND ADAPTIVE DECONVOLUTION AND ITS APPLICATIONS

Blind Deconvolution Using an Improved L0 Sparse Representation

L∞ Blind Deconvolution for the Generalized AR Model.

Identifiability of Multichannel Blind Deconvolution and Nonconvex Regularization Algorithm

Algorithm for Multichannel Blind Deconvolution Based on Subspace Decomposition

Multichannel Sparse Blind Deconvolution on the Sphere

Blind deconvolution using maximum time delay slice of third order moment

Rapid, Robust, and Reliable Blind Deconvolution Via Nonconvex Optimization

Unfolding Neural Networks for Compressive Multichannel Blind Deconvolution

Audio Signal Blind Deconvolution Based on the Quotient Space Hierarchical Theory.

Fast and Provable Simultaneous Blind Super-Resolution and Demixing for Point Source Signals: Scaled Gradient Descent without Regularization

Blind Deconvolution using Convex Programming

Multi-Scale Blind Motion Deblurring Using Local Minimum

A Solution to Blind Separation of Convolutive Communication Mixtures in Frequency Domain

Blindly separating mixtures of multiple layers with spatial shifts

Blind Signal Noise Separation on Instant Mixing Nonlinear Circuits Based on MISEP Algorithm

Blind Image Deconvolution Using the Zero-Lag Slice of Higher-Order Statistics

Generalized variational framework with minimax optimization for parametric blind deconvolution

Identifiability Conditions for Multi-channel Blind Deconvolution with Short Filters