1: \begin{abstract}

2: In this paper, we study the inverse scattering problem for a class of signals

3: that have a compactly supported reflection coefficient. The problem boils down

4: to the solution of the Gelfand-Levitan-Marchenko (GLM) integral equations with a

5: kernel that is bandlimited. By adopting a sampling theory approach

6: to the associated Hankel operators in the Bernstein spaces, a constructive proof

7: of existence of a solution of the GLM equations is obtained under various

8: restrictions on the nonlinear impulse response (NIR). The formalism developed in this

9: article also lends itself well to numerical computations yielding algorithms

10: that are shown to have algebraic rates of convergence. In particular, the use Whittaker-Kotelnikov-Shannon 

11: sampling series yields an algorithm that converges as $\bigO{N^{-1/2}}$ whereas the use

12: of Helms and Thomas (HT) version of the sampling expansion yields an algorithm that

13: converges as $\bigO{N^{-m-1/2}}$ for any $m>0$ provided the regularity

14: conditions are fulfilled. The complexity of the algorithms

15: depend on the linear solver used. The use of conjugate-gradient (CG) method yields an

16: algorithm of complexity $\bigO{N_{\text{iter.}}N^2}$ per sample of the signal 

17: where $N$ is the number of sampling basis functions used and $N_{\text{iter.}}$

18: is the number of CG iterations involved. The HT version of the sampling

19: expansions facilitates the development of algorithms of complexity 

20: $\bigO{N_{\text{iter.}}N\log N}$ (per sample of the signal) by exploiting the special structure as well as the 

21: (approximate) sparsity of the matrices involved. The algorithms are numerically 

22: validated using Schwartz class functions as NIRs that are either bandlimited or

23: effectively bandlimited. The results suggest that the HT variant of our

24: algorithm is spectrally convergent for an input of the aforementioned class.

25: \end{abstract}

26: