1: \begin{abstract}

2:   One-bit compressed sensing is very popular in signal processing and communications

3:   due to its low storage costs and low hardware complexity, but it is a challenging task

4:   to recover the signal by using the one-bit information. In this paper, we propose

5:   a zero-norm regularized smooth difference of convexity (DC) loss model and

6:   derive a family of equivalent nonconvex surrogates covering the MCP and SCAD

7:   surrogates as special cases. Compared to the existing models, the new model and

8:   its SCAD surrogate have better robustness. To compute their $\tau$-stationary points,

9:   we develop a proximal gradient algorithm with extrapolation and establish the convergence

10:   of the whole iterate sequence. Also, the convergence is proved to have a linear rate

11:   under a mild condition by studying the KL property of exponent $0$ of the models.

12:   Numerical comparisons with several state-of-art methods show that

13:   in terms of the quality of solution, the proposed model and its SCAD surrogate

14:   are remarkably superior to the $\ell_p$-norm regularized models,

15:   and are comparable even superior to those sparsity constrained models

16:   with the true sparsity and the sign flip ratio as inputs.

17:  \end{abstract}