Sampling set conditions in weighted multiply generated shift-invariant spaces and their applications (Q2381645)

The authors consider conditions for stable sampling in the spaces \[ V_v^p(\Phi)=\Bigl\{ \sum_{i=1}^r \sum_{k\in\mathbb{Z}^d} c_{ik} \varphi_i (\cdot -k):\, \{c_{ik}\}\in \ell^p_v, i=1,\dots, r \Bigr\} \] where \(\ell^p_v\) consists of those sequences \(\{c_k\}\) such that \(\sum_{k\in \mathbb{Z}^d} | c_k| ^p v(k)\) converges. Here \(v\) is a positive weight function that is assumed to be \(w\)-moderate in the sense that \(0<v(x+y)\leq w(x) v(y)\). One also sets \[ W_0(L^1_\omega) =\left\{f:\| f\| _{W(\omega)}=\sum_{k\in\mathbb{Z}^d} \text{ ess \, sup}_{x\in [0,1)^d}| f(x+k)| \omega(k)<\infty\right\}. \] In addition to the assumption that each \(\varphi_i\in W(L^1_\omega)\) one also assumes that \[ {1\over A}\| f\| _{L^p_v}\leq \sum_{i=1}^r \Biggl\| \Biggl(\int f(x) \varphi_i(x-j) \, dx\Biggr)_{j\in{\mathbb{Z}_d}}\Biggr\| _{\ell^p_v} \leq A\| f\| _{L^p_v} \] for each \(1\leq i\leq r\). Let \(X=\{x_j\} \subset \mathbb{R}^d\) be a discrete set that is \(\gamma\)-dense in the sense that any ball of radius \(\gamma\) contains some element of \(X\) and let weight functions \(\psi_{x_j}\) satisfy, for each \(j\), \(\int \psi_{x_j}=1\), \(\int | \psi_{x_j}| \leq M\) and \(\text{ supp}\, \psi_{x_j}\subset x_j+[-a,a]^d\) for some \(a>0\). Under these hypotheses the authors prove, essentially, the inequality, for any \(f\in V_v^p(\Phi)\), \[ \Bigl(\sum_j | \langle f,\, \psi_{x_j}\rangle| ^p \Bigr)^{1/p}\geq K \| f\| _{L^p_v} \] where \[ K={1\over C_1 V(\gamma,d))^{1/p}} \Bigl(1-BO(\gamma) -MBO(a)(1+O(\gamma))\| _{W(\omega)})\Bigr). \] Here \(V(\gamma,d)\) denotes the volume of a ball of radius \(\gamma\) in \(\mathbb{R}^d\), \(\text{ osc}_\delta (f)(x)=\sup_{| y| <\delta} | f(x+y| -f(x)| \), \(O(\gamma)=\max_{1\leq i\leq r} \| \text{ osc}_\gamma (\varphi_i)\| _{W(\omega)}\), \(C_1\) is such that \(v(j)/C_1 \leq v(x+j)\leq C_1 v(j)\) for all \(j\in \mathbb{Z}^d\) and all \(x\) in a fixed compact set \(K\), and \(B\) is essentially the constant \(A\) in the condition on \(\Phi\) above. A similar but somewhat simpler upper bound is provided on the sum \(\Bigl(\sum_j | \langle f,\, \psi_{x_j}\rangle| ^p \Bigr)^{1/p}\) in terms of \(\| f\| _{L^p_v}\). In this case the sequence \(X\) also has to be uniformly discrete (\(\inf | x_i-x_j| >0\)).