Gabor (super)frames with Hermite functions (Q842303): Difference between revisions

Writing \(\pi_{\lambda}g=e^{2\pi i\lambda_2t}g(t-\lambda_1)\) for the time-frequency shift by \(\lambda=(\lambda_1,\lambda_2)\in{\mathbb R}^2\), the set \[ {\mathcal G}(g,\Lambda)=\{\pi_{\lambda}g\, : \,\lambda\in\Lambda\} \] is called a \textit{Gabor frame} or \textit{Weyl-Heisenberg frame}, when there exist constants \(A,\,B>0\) such that for all \(f\in L^2({\mathbb R})\) \[ A\|f\|^2_{L^2({\mathbb R})}\leq \sum_{\lambda\in\Lambda}\,|\langle f,\pi_{\lambda}g\rangle|^2_{L^2({\mathbb R})} \leq B\|f\|^2_{L^2({\mathbb R})}.\tag{1} \] A \textit{vector-valued} Gabor frame (introduced under the name ``superframe'') is situated in the Hilbert space \({\mathcal H}=L^2({\mathbb R},{\mathbb C}^n)\) of all vector-valued functions \({\mathbf f}(t)=(f_1(t),\dots,f_n(t))\) with inner product \[ \langle{\mathbf f},{\mathbf g}\rangle=\sum_{j=1}^n\,\int_{-\infty}^{\infty}\,f_j(t)\overline{g_j(t)}dt= \sum_{j=1}^n\,\langle f_j,g_j\rangle. \] The \textit{time-frequency shifts} \(\pi_z,\,z=(x,\xi)\) act coordinate wise by \[ \pi_z{\mathbf f}(t)=e^{2\pi i\xi t}{\mathbf f}(t-x). \] The vector-valued Gabor system \({\mathcal G}({\mathbf g},\Lambda)=\{\pi_{\lambda}\text{g}\,:\,\lambda\in\Lambda\}\) is a frame for \(L^2({\mathbb R},{\mathbb C}^n)\) if there exist constants \(A,\,B>0\) such that (1) holds with \(f,\,g\) replaced by \({\mathbf f},\,{\mathbf g}\) and the norm in \(L^2({\mathbb R},{\mathbb C}^n)\) is used. The conditions for a set of functions to form a superframe have been studied extensively (a sufficient condition has for instance been given by \textit{H. Führ} [ Adv. Comput. Math. 29, No. 4, 357--373 (2008; Zbl 1171.42018)] and the main result of the paper is a necessary \textit{and} condition. The Hermite functions \((H_0,\dots,H_n)\) are the unique orthonormal set \(\{f_0,\dots,f_n\}\) in \(L^2({\mathbb R})\) with \(n+1\) elements that satisfies the normalization \[ \int_{-\infty}^{\infty}\,t|f_j(t)|^2dt=0,\;\int_{-\infty}^{\infty}\,\xi |\hat{f}_j(\xi)|^2d\xi=0,\;j=0,\dots,n \] such that the uncertainty \[ \sum_{j=0}^n\,\left(\int_{-\infty}^{\infty}\,t^2|f_j(t)|^2dt+\int_{-\infty}^{\infty}\,\xi^2|\hat{f}_j(\xi)|^2d\xi \right) \] is minimized. Consider now the lattice \(\Lambda=A{\mathbb Z}^2\) for some invertible real-valued \(2\times 2\) matrix \(A\) and let \(s(\Lambda)=|\det{A}|\) be the area of the fundamental domain of \(\Lambda\), then the author proves: Theorem. Let \({\mathbf h}=(H_0,\dots,H_n)\) be the vector of the first \(n+1\) Hermite functions. Then \({\mathcal G}({\mathbf h},\Lambda)\) is a frame for \(L^2({\mathbb R},{\mathbb C}^n)\) if and only if \(s(\Lambda)<{1\over n+1}.\) Finally a result on the lower frame bound is given for \(n=0\): Assume \(s(\Lambda)=1\) and \(0<q<1\). Let \(A_q\) denote the optimal lower frame bound for \({\mathcal G}(H_0,q\Lambda)\), then \[ A_q\geq c(1-q^2)^2=c(1-s(q\Lambda))^2 \] for some constant independent of \(q\).