Consider L groups of point sources or spike trains, with the lth group represented by xl(t). For a function g:R→R, let gl(t)=g(t/μl) denote a point spread function with scale μl>0, and with μ1<⋯<μL. With y(t)=∑Ll=1(gl⋆xl)(t), our goal is to recover the source parameters given samples of y, or given the Fourier samples of y. This problem is a generalization of the usual super-resolution setup wherein L=1; we call this the multi-kernel unmixing super-resolution problem. Assuming access to Fourier samples of y, we derive an algorithm for this problem for estimating the source parameters of each group, along with precise non-asymptotic guarantees. Our approach involves estimating the group parameters sequentially in the order of increasing scale parameters, i.e., from group 1 to L. In particular, the estimation process at stage 1≤l≤L involves (i) carefully sampling the tail of the Fourier transform of y, (ii) a \emph{deflation} step wherein we subtract the contribution of the groups processed thus far from the obtained Fourier samples, and (iii) applying Moitra's modified Matrix Pencil method on a deconvolved version of the samples in (ii).

Citation information

S. Chretien and H. Tyagi, Multi-kernel unmixing and super-resolution using the Modified Matrix Pencil method, 2018, arxiv.org/abs/1807.02862

Research areas