25
2

Universal 1-Bit Compressive Sensing for Bounded Dynamic Range Signals

Abstract

A {\em universal 1-bit compressive sensing (CS)} scheme consists of a measurement matrix AA such that all signals xx belonging to a particular class can be approximately recovered from sign(Ax)\textrm{sign}(Ax). 1-bit CS models extreme quantization effects where only one bit of information is revealed per measurement. We focus on universal support recovery for 1-bit CS in the case of {\em sparse} signals with bounded {\em dynamic range}. Specifically, a vector xRnx \in \mathbb{R}^n is said to have sparsity kk if it has at most kk nonzero entries, and dynamic range RR if the ratio between its largest and smallest nonzero entries is at most RR in magnitude. Our main result shows that if the entries of the measurement matrix AA are i.i.d.~Gaussians, then under mild assumptions on the scaling of kk and RR, the number of measurements needs to be Ω~(Rk3/2)\tilde{\Omega}(Rk^{3/2}) to recover the support of kk-sparse signals with dynamic range RR using 11-bit CS. In addition, we show that a near-matching O(Rk3/2logn)O(R k^{3/2} \log n) upper bound follows as a simple corollary of known results. The k3/2k^{3/2} scaling contrasts with the known lower bound of Ω~(k2logn)\tilde{\Omega}(k^2 \log n) for the number of measurements to recover the support of arbitrary kk-sparse signals.

View on arXiv
Comments on this paper