Wavelet frames for (not necessarily reducing) affine subspaces (Q1026462)

The authors consider affine subspaces of \(L^2(\mathbb{R})\), that is closures of the linear span of a set of the form \(\{2^{n/2}\varphi(2^nt-k): n,k\in \mathbb{Z}, \varphi \in \Phi\}\) where \(\Phi\) is a finite set of functions. A subspace is said to be a reducing subspace if it is invariant under dilations \((Df)(t)= \sqrt 2f(2t)\) (which is automatically the case for affine subspaces) and under translations \((Tf)(t)=f(t-1)\). For a reducing affine subspace \(X\) it is known that there is a functions such that \(\{D^nT^k\psi: n,k\in \mathbb{Z}\}\) is an orthonormal basis for \(X\) and one of the main results of this paper is that if \(X\) is an affine subspace generated by a single function, i.e., \(\Phi=\{\varphi\}\), then there are two functions \(\psi_1\) and \(\psi_2\) so that \(\{D^nT^k\psi_i: n,k\in \mathbb{Z}, i=1,2\}\) is a Parseval frame for \(X\), i.e. \(\|f\|^2 = \sum_{n,k,i}|\langle f,D^nT^k\psi_i\rangle|^2 \) for all \(f\in X\). In addition several sufficient conditions for the existence of one function \(\psi\) so that \(\{D^nT^k\psi: n,k\in \mathbb{Z} \}\) is a Parseval frame for \(X\) are given.