1: \begin{abstract}

2: This paper proposes a fast approximate message-passing (AMP)

3: algorithm for solving compressed sensing (CS) recovery problems with

4: 1D-finite-difference  sparsity in term of MMSE estimation. The

5: proposed algorithm, named ssAMP-BGFD, is low-computational with its

6: fast convergence and cheap per-iteration cost, providing phase

7: transition nearly approaching to the state-of-the-art. The proposed

8: algorithm is originated from a sum-product message-passing rule,

9: applying a Bernoulli-Gaussian (BG) prior, seeking an MMSE solution.

10: The algorithm construction includes not only the conventional AMP

11: technique for the measurement fidelity, but also suggests a

12: simplified message-passing method to promote the signal sparsity in

13: finite-difference. Furthermore, we provide an EM-tuning methodology

14: to learn the BG prior parameters, suggesting how to use some

15: practical measurement matrices satisfying the RIP requirement under

16: the ssAMP-BGFD recovery. Extensive empirical results confirms

17: performance of the proposed algorithm, in phase transition,

18: convergence speed, and CPU runtime, compared to the recent

19: algorithms.

20: \end{abstract}

21: