Abstract
Reproducing systems in \(L^2({\mathbb {R}})\) such as wavelet and Gabor dual frames have been extensively studied, but reducing systems in \(L^2({\mathbb {R}}_+)\) with \({{\mathbb {R}}_+}=(0,\,\infty )\) have not. In practice, \(L^2({\mathbb {R}}_+)\) models the causal space since the time variable cannot be negative. Due to \({\mathbb {R}}_+\) not being a group under addition, \(L^2({\mathbb {R}}_+)\) admits no nontrivial shift invariant system and thus admits no traditional wavelet or Gabor analysis. However, \(L^2({\mathbb {R}}_+)\) admits nontrivial dilation systems due to \({\mathbb {R}}_+\) being a group under multiplication. This paper addresses the frame theory of a class of dilation-and-modulation (\({\mathcal{M}\mathcal{D}}\)) systems generated by a finite family in \(L^2({\mathbb {R}}_+)\). We obtain a parametric expression of \({\mathcal{M}\mathcal{D}}\)-frames, and a density theorem for such \({\mathcal{M}\mathcal{D}}\)-systems which is parallel to that of traditional Gabor systems in \(L^2({\mathbb {R}})\). It is well known that an arbitrary Gabor frame must admit dual frames with the same structure. Interestingly, it is not the case for \({\mathcal{M}\mathcal{D}}\)-frames. We prove that an \({\mathcal{M}\mathcal{D}}\)-frame admits \({\mathcal{M}\mathcal{D}}\)-dual frames if and only if \(\log _ba\) is an integer, where a and b are dilation and modulation parameters, respectively. And in this case, we characterize and express all \({\mathcal{M}\mathcal{D}}\)-dual generators for an arbitrarily given \({\mathcal{M}\mathcal{D}}\)-frame. Some examples are also provided.
Similar content being viewed by others
Avoid common mistakes on your manuscript.
1 Introduction
Recall that a countable family \(\{f_{j}\}_{j\in J}\) in a separable Hilbert space \({{\mathcal {H}}}\) is called a frame for \({{\mathcal {H}}}\) if there exist constants \(0<A\le B<\infty \) satisfying
where A and B are called frame bounds. If only the right-hand side inequality holds, we say that \(\{f_{j}\}_{j\in J}\) is a Bessel sequence with Bessel bound B. And we say that it is a tight frame (Parseval frame) for \({{\mathcal {H}}}\) if \(A=B\) (\(A=B=1\)) in (1.1). In addition, \(\{f_{j}\}_{j\in J}\) is said to be a frame sequence in \({{\mathcal {H}}}\) if it is a frame for its closed linear span \(\overline{\textrm{span}}(\{f_{j}\}_{j\in J})\). And it is called a Riesz sequence in \({\mathcal {H}}\) if there exist constants \(0<A\le B<\infty \) satisfying
where \(l_{0}(J)\) denotes the set of all finitely supported sequences on J, A and B are called Riesz bounds. And it is called a Riesz basis for \({\mathcal {H}}\) if it is a Riesz sequence and complete. Given a frame \(\{f_{j}\}_{j\in J}\) for \({{\mathcal {H}}}\), a sequence \(\{g_{j}\}_{j\in J}\) in \({{\mathcal {H}}}\) is called a dual frame of \(\{f_{j}\}_{j\in J}\) if it is a frame for \({{\mathcal {H}}}\) satisfying
By a standard argument, we have that \(\{f_{j}\}_{j\in J}\) is also a dual of \(\{g_{j}\}_{j\in J}\) if \(\{g_{j}\}_{j\in J}\) is a dual of \(\{f_{j}\}_{j\in J}\) (so, in this case, we say \(\{f_{j}\}_{j\in J}\) and \(\{g_{j}\}_{j\in J}\) form a pair of dual frames for \({\mathcal {H}}\)), and that \(\{f_{j}\}_{j\in J}\) and \(\{g_{j}\}_{j\in J}\) form a pair of dual frames for \({\mathcal {H}}\) if they are Bessel sequences satisfying (1.3). We refer to [3, 12, 16, 32] for basics on frames. Equation (1.3) gives a reproducing formula for vectors in \({{\mathcal {H}}}\), where \(\{g_{j}\}_{j\in J}\) is used for analyzing f and \(\{f_{j}\}_{j\in J}\) for reconstructing f. So \(\{f_{j}\}_{j\in J}\) and \(\{g_{j}\}_{j\in J}\) satisfying (1.3) are also said to form a reproducing system. Dual frame pairs are stable reproducing systems. During the past more than 30 years, reproducing systems such as wavelet and Gabor dual frames for \(L^2({\mathbb {R}}_+)\) have been extensively studied ( [2, 4,5,6,7, 9,10,11, 13, 17, 19,20,22, 29, 31]). But the theory of structured frames for \(L^2({\mathbb {R}}_+)\) has not. It is because \({\mathbb {R}}\) is a group while \({\mathbb {R}}_+\) is not under addition. This results in \(L^2({\mathbb {R}}_+)\) with \({{\mathbb {R}}_+}=(0,\,\infty )\) admitting no nontrivial shift invariant system and thus admitting no traditional wavelet or Gabor analysis. Observe that \({\mathbb {R}}_+\) is a group under multiplication. This makes \(L^2({\mathbb {R}}_+)\) to be closed under dilation. This paper addresses a class of dilation-and-modulation reproducing systems generated by a finite family in \(L^2({\mathbb {R}}_+)\).
For the moment, in order to explain the scope of this paper, we recall some notions and notations. \({\mathbb {Z}}\) and \({\mathbb {N}}\) denote the set of integers and the set of positive integers, respectively. Given a, \(b>1\), a measurable function h defined on \({\mathbb {R}}_+\) is said to be b-dilation periodic if \(h(b\cdot )=h(\cdot )\) a.e. on \({\mathbb {R}}_+\). Obviously, a b-dilation periodic function on \({\mathbb {R}}_+\) is determined by its values on \([1,\,b)\). \(\{\Lambda _m\}_{m\in {\mathbb {Z}}}\) denotes the sequence of b-dilation periodic functions on \({\mathbb {R}}_+\) defined by
Given a finite family \(\Psi =\{\psi _1,\,\psi _2,\,\ldots ,\, \psi _L\}\) in \(L^2({\mathbb {R}}_+)\), \({{\mathcal{M}\mathcal{D}}}(\Psi ,\,a,\,b)\) denotes the dilation-and-modulation system generated by \(\Psi \):
where
For simplicity, we write \({{\mathcal{M}\mathcal{D}}}(\psi ,a,b)\) for \({{\mathcal{M}\mathcal{D}}}(\Psi ,\,a,\,b)\) if \(\Psi \) is a singleton \(\{\psi \}\), and write \({{\mathcal{M}\mathcal{D}}}(\Psi ,\, a)\) for \({{\mathcal{M}\mathcal{D}}}(\Psi ,\,a,\,b)\) if \(a=b\). In this paper, we work under the following general setup:
General setup. (i) \(a,\,b\) are two constants greater than 1 such that
for some coprime positive integers p and q.
(ii) \(\Psi =\{\psi _1,\,\psi _2,\,\ldots ,\, \psi _L\}\) is a finite family in \(L^2({\mathbb {R}}_+)\).
From (1.6), we have \(b^p=a^q\). Throughout this paper, we always write
In practice, the time variable cannot be negative, and \(L^2({\mathbb {R}}_+)\) models the causal space. \({\mathcal{M}\mathcal{D}}\)-systems of the form (1.5) have potential applications in analyzing causal signals. As pointed out in [24, 25], an \({\mathcal{M}\mathcal{D}}\)-system is different from the Fourier transform version of a wavelet system in the Hardy space \(H^2({\mathbb {R}})\), and \(L^2({\mathbb {R}}_+)\) is not closed under the Fourier transform. So the Fourier transform is not applicable to analysis for \(L^2({\mathbb {R}}_+)\) although it is a powerful tool for wavelet and Gabor analysis in \(L^2({\mathbb {R}})\). Some special cases of (1.5) were studied in [23,24,25,26,27, 30]. [23, 25] investigated \({{\mathcal{M}\mathcal{D}}}(\Psi ,\,a)\)-frames with \(\Psi \) being a finite family of characteristic functions, where [25] is for \(\textrm{card}(\Psi )=1\) and [23] is for general \(\textrm{card}(\Psi )\). [27] characterized \({{\mathcal{M}\mathcal{D}}}(\Psi ,\,a)\)-frames and dual frames, and [30] is for the vector-valued case. [24] presented a density theorem for singly generated \({{\mathcal{M}\mathcal{D}}}\)-system of the form \({{\mathcal{M}\mathcal{D}}}(\psi ,a,b)\). This paper focuses on \({\mathcal{M}\mathcal{D}}\)-dual pairs of the form (1.5) under the general setup. Two frames \({{\mathcal{M}\mathcal{D}}}(\Psi ,\,a,\,b)\) and \({{\mathcal{M}\mathcal{D}}}(\Phi ,\,a,\,b)\) for \(L^2({\mathbb {R}}_+)\) form a pair of dual frames if
equivalently,
Given Bessel sequences \({{\mathcal{M}\mathcal{D}}}(\Psi ,\,a,\,b)\) and \({{\mathcal{M}\mathcal{D}}}(\Phi ,\,a,\,b)\) in \(L^{2}({\mathbb {R}}_+)\), define the mixed frame operator \(S_{\Psi ,\,\Phi }\) by
Then, it is a bounded operator on \(L^{2}({\mathbb {R}}_+)\). And (1.8) and (1.9) may be written as \(S_{\Phi ,\,\Psi }=I\) and \(S_{\Psi ,\,\Phi }=I\), respectively, where I is the identity operator on \(L^2({\mathbb {R}}_+)\). Write \(S_{\Psi ,\Psi }=S_{\Psi }\).
Recall that an arbitrary Gabor frame for \(L^2({\mathbb {R}})\) admits dual frames with Gabor structure. In contrast, this paper points out that not every \({\mathcal{M}\mathcal{D}}\)-frame has the duals with the same structure. We prove that \(\log _ba\in {\mathbb {N}}\) is the necessary and sufficient condition for an \({\mathcal{M}\mathcal{D}}\)-frame \({{\mathcal{M}\mathcal{D}}}(\Psi ,\,a,\,b)\) admitting \({\mathcal{M}\mathcal{D}}\)-dual frames (see Theorem 3.1), and present a parametric expression of \({\mathcal{M}\mathcal{D}}\)-frames and their \({\mathcal{M}\mathcal{D}}\)-duals (see Theorems 2.3, 3.2 and 3.4). It is one of the novelties of this paper. Another feature is our method of proof. Recall that the traditional Fourier transform is not applicable to analysis for \(L^2({\mathbb {R}}_+)\). We introduce the \(\Theta _\beta \)-transform matrix methods, and, with the help of dilation congruence and disjointness, reduce all problems to designing suitable \(\Theta _\beta \)-transform matrix-valued functions on \([1,\,a^{\frac{1}{p}})\times [0,\,1)\) of finite order. All entries of such matrices are mutually independent. This gives us much freedom in designing the \(\Theta _\beta \)-transform matrix-valued functions and makes all conditions of theorems easily realized. Of course, our general setup is a technical requirement. How to treat the case of \(\log _ba\) being irrational is unresolved.
The rest of this paper is organized as follows. Section 2 is devoted to parametric expression of \({\mathcal{M}\mathcal{D}}\)-Bessel sequences (frames and Riesz bases) and the density theorem of \({\mathcal{M}\mathcal{D}}\)-systems. Section 3 focuses on proving \(\log _ba\in {\mathbb {N}}\) being the necessary and sufficient condition for an \({\mathcal{M}\mathcal{D}}\)-frame to admit \({\mathcal{M}\mathcal{D}}\)-duals, and a parametric expression of \({\mathcal{M}\mathcal{D}}\)-duals. Some examples are also provided.
Before proceeding, we introduce some notations and notions. Throughout this paper, inclusion or equality between two measurable sets in \({\mathbb {R}}_+\) (equality or inequality between two functions) means it holding up to a set of measure zero. Given M, \(N\in {\mathbb {N}}\), we denote by \(I_M\) the \(M\times M\) identity matrix, by \({{\mathbb {N}}}_M\) the set \(\{0,\,1,\,2,\,\ldots ,M-1\}\), by \({{\mathcal {M}}}_{M,\,N}\) the set of \(M\times N\) complex matrices. For \({{\mathcal {A}}}=({{\mathcal {A}}}_{m,n})_{1\le m\le M,\,1\le n\le N}\in {{\mathcal {M}}}_{M,\,N}\), we denote by \(\Vert {{\mathcal {A}}}\Vert _2\) its Frobenius norm, i.e.,
Obviously, \(\Vert {{\mathcal {A}}}\Vert _2=\Vert {{\mathcal {A}}}\Vert _{{{\mathbb {C}}}^M}\) if \(N=1\). We denote by \({{\mathcal {A}}}^t\) and \({{\mathcal {A}}}^*\) its transpose and conjugate transpose, respectively, by \(I_L\otimes {{\mathcal {A}}}\) the block matrix (with L blocks) of the form \(I_L\otimes {{\mathcal {A}}}=\textrm{diag}({{\mathcal {A}}},\,{{\mathcal {A}}},\ldots ,\,{{\mathcal {A}}})\), by \(\{e_{m,\,j}\}\) the function sequence:
where \(\Lambda _m\) is as in (1.4); and by \(\{e_k:\,k\in {{\mathbb {N}}}_M\}\) denotes the canonical orthonormal basis for \({{\mathbb {C}}}^M\), i.e., each \(e_k\) is the vector with the kth component being 1 and the others being 0.
Definition 1.1
Let a, b and \(\Psi \) be as in the general setup, and \(\beta \) be as in (1.7). Define \(\Theta _\beta :\,L^{2}({\mathbb {R}}_+)\rightarrow L^{2}_{loc}({{\mathbb {R}}_+}\times {{\mathbb {R}}})\), \(\Gamma :\,L^{2}({\mathbb {R}}_+)\rightarrow L^{2}_{loc}({{\mathbb {R}}_+}\times {{\mathbb {R}}},\,{{\mathbb {C}}}^p)\) and \(\Upsilon :\,L^{2}({\mathbb {R}}_+)\rightarrow L^{2}_{loc}({{\mathbb {R}}_+}\times {{\mathbb {R}}},\,{{\mathcal {M}}}_{q,\,p})\) by
for \(f\in L^2({\mathbb {R}}_+)\) and a.e. \((x,\,\xi )\in {{\mathbb {R}}_+}\times {\mathbb {R}}\), respectively. We associate \(\Psi \) with the matrix-valued function (so-called \(\Theta _{\beta }\)-transform matrix) from \({\mathbb {R}}_+\times {\mathbb {R}}\) into \({{\mathcal {M}}}_{Lq,\,p}\) defined by
for a.e. \((x,\xi )\in {{\mathbb {R}}_+}\times {{\mathbb {R}}}\).
Remark 1.1
(i) The \(\Theta _{\beta }\)-transform herein appeared before in [14] called multiplicative Zak transform and [18] called generalized Zak transform. A more general definition appeared in [1] to study reproducing systems on LCA groups.
(ii) By ([24], Remark 2.4) and a standard argument, \(\Theta _{\beta }\), \(\Gamma \) and \(\Upsilon \) are all well defined. And by Lemma 2.2 (v), \(\Psi \) is uniquely determined by the values of its \(\Theta _{\beta }\)-transform matrix \({\varvec{\Psi }}(x,\,\xi )\) for \((x,\,\xi )\in E_{\Upsilon }\times [0,\,1)\) where \(E_{\Upsilon }\) is \(({a^{\frac{1}{p}}})^{{\mathbb {Z}}}\)-dilation congruent to \([1,\,a^{\frac{1}{p}})\). For an arbitrary \(Lq\times p\) matrix-valued function defined on \(E_{\Upsilon }\times [0,\,1)\) with entries in \(L^{2}(E_{\Upsilon }\times [0,\,1))\) determines a unique \(\Psi \) in \(L^2({\mathbb {R}}_+)\) via (1.14) restricted to \(E_{\Upsilon }\times [0,1)\). And the unique \(\Psi =\{\psi _1,\,\psi _2,\,\ldots ,\,\psi _L\}\) is defined by
for \(1\le l\le L\).
Definition 1.2
For a measurable set \(S\subset {\mathbb {R}}_+\), a collection \(\{S_i\}_{i\in {\mathcal {I}}}\) of measurable subsets of S is called a partition of S if \(\chi _{_S}=\sum \limits _{i\in {\mathcal {I}}}\chi _{_{S_i}}\), where \(\chi _{_E}\) denotes the characteristic function of E for a set E. And for \(\alpha >0\), measurable subsets T, \({\widetilde{T}}\) of \({\mathbb {R}}_+\) and a collection \(\{\Gamma _i:\,i\in {\mathcal {I}}\}\) of measurable subsets of \({\mathbb {R}}_+\) with \({\mathcal {I}}\) being at most countable, we say that T is \({\alpha }^{{\mathbb {Z}}}\)-dilation congruent to \({\widetilde{T}}\) if there exists a partition \(\{T_k:\,k\in {\mathbb {Z}}\}\) of T such that \(\{\alpha ^kT_k:\,k\in {\mathbb {Z}}\}\) is a partition of \({\widetilde{T}}\), and that \(\{\Gamma _i:\,i\in {{\mathcal {I}}}\}\) is \({\alpha ^{{\mathbb {Z}}}}\)-dilation disjoint if
for \(i,\,j\in {{\mathcal {I}}}\) and \(k\in {\mathbb {Z}}\) with \((i,\,k)\ne (j,\,0)\).
Remark 1.2
By Definition 1.2, \({\widetilde{T}}\) is \(\alpha ^{{\mathbb {Z}}}\)-dilation congruent to T if T is \(\alpha ^{{\mathbb {Z}}}\)-dilation congruent to \(\widetilde{T}\). So we say that T and \({{\widetilde{T}}}\) are \(\alpha ^{{\mathbb {Z}}}\)-dilation congruent. Also observe that only finitely many \(T_{k}\) among \(\{T_{k}: \,k\in {\mathbb {Z}}\}\) are nonempty if T and \({\widetilde{T}}\) are contained in some bounded subinterval \([M,\,N]\) of \({\mathbb {R}}_+\).
Definition 1.3
Let a, b, \(\Psi \) be as in the general setup, and \({{\mathcal{M}\mathcal{D}}}(\Psi ,\,a,\,b)\) be a Bessel sequence in \(L^2({\mathbb {R}}_+)\). We say \({{\mathcal{M}\mathcal{D}}}(\Psi ,\,a,\,b)\) has Riesz property if \(c=0\) is the unique solution to
in \(l^2(\{1,2,\ldots , L\}\times {{\mathbb {Z}}^2})\).
2 Parametric expression and density theorem
This section is devoted to parametric expression of \({\mathcal{M}\mathcal{D}}\)-Bessel sequences (frames and Riesz bases) and the density theorem of \({\mathcal{M}\mathcal{D}}\)-systems (see Theorems 2.3 and 2.4). For this purpose, we need to establish some lemmas. The first one gives a dilation congruence-based partition of \({\mathbb {R}}_+\).
Lemma 2.1
Given \(\alpha >1\), let E be \(\alpha ^{{\mathbb {Z}}}\)-dilation congruent to \([1,\,\alpha )\). Then, \(\{\alpha ^jE:\,j\in {\mathbb {Z}}\}\) is a partition of \({{\mathbb {R}}}_+\).
Proof
Suppose \(\{E_k:\,k\in {\mathbb {Z}}\}\) is a partition of E such that \(\{{\alpha }^kE_k:\,k\in {\mathbb {Z}}\}\) is a partition of \([1,\,\alpha )\). Also observe that \(\{{\alpha }^j[1,\,\alpha ):\,j\in {\mathbb {Z}}\}\) is a partition of \({{\mathbb {R}}}_+\). It follows that \(\{{\alpha }^{k+j}E_k:\,j,\,k\in {\mathbb {Z}}\}\) is a partition of \({{\mathbb {R}}}_+\). Thus,
Interchanging the order of summation gives
where we used the fact that \(\{E_k:\,k\in {\mathbb {Z}}\}\) is a partition of E in the last equality. The proof is completed. \(\square \)
Lemma 2.2
Let \(a,\,b\) be as in the general setup. Then,
(i) \(\{e_{m,\,j}:\,m,\,j\in {\mathbb {Z}}\}\) is an orthonormal basis for \(L^2([1,\,b)\times [0,\,1))\).
(ii)
for \(f\in L^1([1,\,b)\times [0,\,1))\).
(iii) \(\Theta _{\beta }\) has the quasi-periodicity property:
for a.e. \((x,\,\xi )\in {{\mathbb {R}}}_+\times {{\mathbb {R}}}\).
(iv) For \(f\in L^2({\mathbb {R}}_+)\) and \((j,\,m,\,r)\in {{\mathbb {Z}}}\times {{\mathbb {Z}}}\times {{\mathbb {N}}}_q\),
for a.e. \((x,\,\xi )\in {{\mathbb {R}}}_+\times {{\mathbb {R}}}\).
(v) The mapping \(\Theta _\beta \), \(\Gamma \) and \(\Upsilon \) are unitary operators from \(L^2({\mathbb {R}}_+)\) onto \(L^2(E_{\Theta }\times [0,\,1))\), \(L^2(E_{\Gamma }\times [0,\,1),\,{{\mathbb {C}}}^p)\) and \(L^2(E_{\Upsilon }\times [0,\,1),\,{{\mathcal {M}} }_{q,p})\), respectively, where \(E_{\Theta }\) is \(\beta ^{{\mathbb {Z}}}\)-dilation congruent to \([1,\,\beta )\), \(E_{\Gamma }\) is \(b^{{\mathbb {Z}}}\)-dilation congruent to \([1,\,b)\) and \(E_{\Upsilon }\) is \(({a^{\frac{1}{p}}})^{{\mathbb {Z}}}\)-dilation congruent to \([1,\,a^{\frac{1}{p}})\).
Proof
(i)–(iv) are repeated from ([24], Lemmas 3.1,2.2). Next we prove (v). Obviously \(\Theta _\beta \), \(\Gamma \) and \(\Upsilon \) are all linear. We only need to prove that they are norm-preserving and onto. Fix an arbitrary \(f\in L^2({\mathbb {R}}_+)\). We have
where the last equality follows from the fact that \(\{e^{-2\pi ik \cdot }:\,k\in {\mathbb {Z}}\}\) is an orthonormal basis for \(L^2([0,\,1))\). Since \(E_{\Theta }\) is \(\beta ^{{\mathbb {Z}}}\)-dilation congruent to \([1,\,\beta )\), there exists a partition of \(\{E_j:\,j\in {\mathbb {Z}}\}\) of \(E_{\Theta }\) such that \(\{\beta ^jE_j:\,j\in {\mathbb {Z}}\}\) is a partition of \([1,\,\beta ).\) It follows that
This implies that \(\Theta _\beta \) is norm-preserving. Next, we prove that it is onto. Let \(F\in L^2(E_{\Theta }\times [0,\,1))\). Define f on \({\mathbb {R}}_+\) by
Since \(E_{\Theta }\) is \(\beta ^{{\mathbb {Z}}}\)-dilation congruent to \([1,\,\beta )\), \(\{\beta ^jE_{\Theta }:j\in {\mathbb {Z}}\}\) is a partition of \({\mathbb {R}}_+\) by Lemma 2.1. It follows that f is well defined on \({\mathbb {R}}_+\), and
And by a standard argument, we have
for a.e. \((x,\,\xi )\in E_{\Theta }\times [0,\,1)\). Hence, \(\Theta _\beta \) is onto.
Now we prove that \(\Gamma \) is norm-preserving and onto. Fix an arbitrary \(f\in L^2({\mathbb {R}}_+)\). Then,
Since \(E_{\Gamma }\) is \(b^{{\mathbb {Z}}}\)-dilation congruent to \([1,\,b)\), \(\{b^kE_{\Gamma }:\,k\in {\mathbb {Z}}\}\) is a partition of \({\mathbb {R}}_+\) by Lemma 2.1. Thus,
This implies that \(\Gamma \) is norm-preserving. Next, we prove that it is onto. Let \(F=\left( F_s\right) _{s\in {{\mathbb {N}}}_p}\in L^2(E_{\Gamma }\times [0,\,1),\,{{\mathbb {C}}}^p)\). Define f on \({\mathbb {R}}_+\) by
By the same procedure as above, we may prove \(f\in L^2({\mathbb {R}}_+)\) and \(\Gamma f=F\). Hence, \(\Gamma \) is onto.
Finally, we prove that \(\Upsilon \) is norm-preserving and onto. Fix an arbitrary \(f\in L^2({\mathbb {R}}_+)\). Then,
where the third equality uses the fact that \({\mathbb {Z}}\) is the disjoint union of \(pr+qs+pq{{\mathbb {Z}}}\) with \(r\in {{\mathbb {N}}_q}\) and \(s\in {{\mathbb {N}}_p}\). Since \(E_{\Upsilon }\) is \((a^{\frac{1}{p}})^{{\mathbb {Z}}}\)-dilation congruent to \([1,\,a^{\frac{1}{p}})\), \(\big \{a^{\frac{k}{p}}E_{\Upsilon }:\,k\in {\mathbb {Z}}\big \}\) is a partition of \({\mathbb {R}}_+\) by Lemma 2.1. Thus,
This implies that \(\Upsilon \) is norm-preserving. Next, we prove that it is onto. Let \(F=\left( F_{r,\,s}\right) _{r\in {{\mathbb {N}}}_q,\,s\in {{\mathbb {N}}}_p}\in L^2(E_{\Upsilon }\times [0,\,1),\,{{\mathcal {M}}}_{q,p})\). Define f on \({\mathbb {R}}_+\) by
By the same procedure as above, we may prove \(f\in L^2({\mathbb {R}}_+)\) and \(\Upsilon f=F\). Hence, \(\Upsilon \) is onto. The proof is completed. \(\square \)
By ([24], Lemma 3.5) and a simple computation, we have the following lemma.
Lemma 2.3
Let \(a,\,b\) and \(\Psi \) be as in the general setup. Then, we have
(i) For \((r,\,j)\in {{\mathbb {N}}}_q\times {\mathbb {Z}}\),
for a.e. \((x,\,\xi )\in {{\mathbb {R}}_+}\times {\mathbb {R}}\), where
(ii)
for a.e. \((x,\,\xi )\in {{\mathbb {R}}_+}\times {\mathbb {R}}\), where \((r',\,s')\in ({{\mathbb {N}}}_q\backslash \{0\})\times ({{\mathbb {N}}}_p\backslash \{0\})\) satisfying \(pr'+qs'=pq+1\),
and
(iii) For \((s,\,j)\in {{\mathbb {N}}}_p\times {\mathbb {Z}}\) and \(f\in L^2({{\mathbb {R}}}_+)\),
and
for a.e. \((x,\,\xi )\in {{\mathbb {R}}_+}\times {\mathbb {R}}\), where
The following lemma is a generalization of ([24], Lemma 3.6).
Lemma 2.4
Let \(a,\,b\) and \(\Psi \) be as in the general setup. Then, for positive constants A and B with \(A\le B\), we have
(i)
if and only if
(ii)
if and only if
Proof
We only prove (i), and (ii) can be proved similarly. Since \(b=a^{\frac{q}{p}}\) and \([1,\,b)=\bigcup \limits _{r\in {{\mathbb {N}}}_q}a^{\frac{r}{p}}[1,\,a^\frac{1}{p})\), (2.8) holds if and only if
for \(r\in {{\mathbb {N}}}_q\) and a.e. \((x,\,\xi )\in [1,\,a^{\frac{1}{p}})\times [0,\,1)\). This can be rewritten as
for \(r\in {{\mathbb {N}}}_q\) and a.e. \((x,\,\xi )\in [1,\,a^{\frac{1}{p}})\times [0,\,1)\) by Lemma 2.3 (ii). It is in turn equivalent to (2.9) by a standard argument. The proof is completed. \(\square \)
Remark 2.1
Lemma 2.4 implies that
is necessary for (2.8) or (2.9) to hold. In particular, \(q=1\) if \(A=B\) in Lemma 2.4.
By ([24], Lemma 4.1), under the general setup, we have
Again by Lemma 2.2 (v), this implies that \({{\mathcal{M}\mathcal{D}}}(\Psi ,\,a,\,b)\) is a Bessel sequence in \(L^2({\mathbb {R}}_+)\) with bound B if and only if
The following lemma reduces (2.14) to the spectral condition of \({\varvec{\Psi }}^*(x,\,\xi ){\varvec{\Psi }}(x,\,\xi )\) on \([1,\,a^{\frac{1}{p}})\times [0,\,1)\).
Lemma 2.5
Let \(a,\,b\) and \(\Psi \) be as in the general setup. Then, \({{\mathcal{M}\mathcal{D}}}(\Psi ,\,a,\,b)\) is a Bessel sequence in \(L^{2}({\mathbb {R}}_+)\) with bound B if and only if
Proof
By Lemma 2.4, (2.15) is equivalent to
Next we prove that \({{\mathcal{M}\mathcal{D}}}(\Psi ,\,a,\,b)\) is a Bessel sequence in \(L^{2}({\mathbb {R}}_+)\) with bound B if and only if (2.16) holds. Observe that
for \(f\in L^2({\mathbb {R}}_+)\). We have that (2.14) holds, and thus \({{\mathcal{M}\mathcal{D}}}(\Psi ,\,a,\,b)\) is a Bessel sequence in \(L^2({\mathbb {R}}_+)\) with bound B if (2.16) holds.
Now we prove the converse implication. Suppose \({{\mathcal{M}\mathcal{D}}}(\Psi ,\,a,\,b)\) is a Bessel sequence in \(L^2({\mathbb {R}}_+)\) with bound B. Then, (2.14) holds. Fix an arbitrary \(c\in {{\mathbb {C}}}^p\) and \(E\subset [1,\,b)\times [0,\,1)\) with \(|E|>0\). Take f in (2.14) by
Then, f is well defined by Lemma 2.2 (v), and
This implies that
by the arbitrariness of E and ([28], Theorem 1.40), equivalently, (2.16) holds. The proof is completed. \(\square \)
Remark 2.2
Observe that (2.15) is equivalent to
Since all norms on an arbitrarily finite-dimensional normed linear spaces are mutually equivalent, \({{\mathcal{M}\mathcal{D}}}(\Psi ,\,a,\,b )\) is a Bessel sequence in \(L^2({\mathbb {R}}_+)\) if and only if all entries of \({\varvec{\Psi }}(x,\xi )\) belong to \(L^{\infty }([1,\,a^{\frac{1}{p}})\times [0,\,1))\), equivalently, they belong to \(L^{\infty }({{\mathbb {R}}}_+\times {{\mathbb {R}}})\) by ([24], Lemma 3.3) and quasi-periodicity of \(\Theta _{\beta }\)-transform.
The following lemma is a partial generalization of (2.13).
Lemma 2.6
Let \(a,\,b\) and \(\Psi \) be as in the general setup, and \(\Phi =\{\phi _1,\,\phi _2,\,\ldots ,\,\phi _L\}\) be a sequence in \(L^2({\mathbb {R}}_+)\). Then,
for \(f,\,g\) satisfying
Proof
Arbitrarily fix \(f,\,g\) satisfying (2.18). By (2.13) and (2.18),
So the left-hand side of (2.17) is well defined, and
Applying Lemma 2.2 to (2.20) gives
Also observe that
by (2.18). It follows that
Therefore, (2.17) holds. The proof is completed. \(\square \)
The following two lemmas are repeated from ([24], Lemmas 4.2, 4.3). The first one is a variation of ([15], Corollary 2.4), and the second one is obtained by an application of the spectral theorem of self-adjoint matrices (see [8], p.978).
Lemma 2.7
An arbitrary \(\mu \times \nu \) matrix-valued measurable function \({{\mathcal {A}}}(\cdot )\) on a measurable set E in \({{\mathbb {R}}}^d\) must have the form
where \(U(\cdot )\) and \(V(\cdot )\) are \(\mu \times \mu \) and \(\nu \times \nu \) unitary matrix-valued measurable functions on E, respectively, and \(\Lambda (\cdot )\) is a \(\min (\mu ,\,\nu )\times \min (\mu ,\,\nu )\) diagonal matrix-valued measurable function on E.
Lemma 2.8
Let \({{\mathcal {A}}}(\cdot )\) be an \(m\times n\) matrix-valued measurable function defined on a measurable set E. Then, the orthogonal projection operator \(P_{ker({\mathcal {A}}(\cdot ))}\) from \({\mathbb {C}}^n\) onto the kernel space \(ker({\mathcal {A}}(\cdot ))\) of \({{\mathcal {A}}}(\cdot )\) is measurable on E, and
The following lemma characterizes the Riesz property of \({\mathcal{M}\mathcal{D}}\)-Bessel sequences.
Lemma 2.9
Let \(a,\,b\) and \(\Psi \) be as in the general setup. Suppose \({{\mathcal{M}\mathcal{D}}}(\Psi ,\,a,\,b)\) is a Bessel sequence in \(L^2({\mathbb {R}}_+)\). Then, the following are equivalent:
(i) \({{\mathcal{M}\mathcal{D}}}(\Psi ,\,a,\,b)\) has Riesz property.
(ii) \(\textrm{rank}({\varvec{\Psi }})(x,\,\xi )=Lq{ \text{ for } \text{ a.e. } }(x,\,\xi )\in [1,\,a^\frac{1}{p})\times [0,\,1).\)
Proof
Since \({{\mathcal{M}\mathcal{D}}}(\Psi ,\,a,\,b)\) is a Bessel sequence in \(L^2({\mathbb {R}}_+)\),
is well defined and converges unconditionally for \(c\in l^2(\{1,\,2,\ldots ,\,L\}\times {{\mathbb {Z}}^2})\). Thus,
if and only if
for some \(c\in l^2(\{1,\,2,\ldots ,\,L\}\times {{\mathbb {Z}}^2})\). By Lemma 2.2 (v), (2.23) is equivalent to
equivalently,
for a.e. \((x,\,\xi )\in [1,\,b)\times [0,\,1)\), i.e.,
where
Thus, (i) is equivalent to \(C(x,\xi )=0\) is a unique solution to (2.24) in \(L^2([1,b)\times [0,1),\,{{\mathbb {C}}}^{Lq})\). Observe
if and only if
by Lemma 2.3 (ii) and the fact that \(b=a^{\frac{q}{p}}\). So we only need to prove that \(C(x,\xi )=0\) being a unique solution to (2.24) in \(L^2([1,b)\times [0,1),\,{{\mathbb {C}}}^{Lq})\) if and only if (2.26) holds.
The sufficiency is obvious. Now, we prove the necessity by proving that (2.24) has a nonzero solution in \(L^2([1,\,b)\times [0,\,1),\,{{\mathbb {C}}}^{Lq})\) if \(\textrm{rank}({\varvec{\Psi }})(\cdot ,\,\cdot )<Lq\) on some \(E\subset [1,\,b)\times [0,\,1)\) with \(|E|>0\). Let \(\{e_k:\,k\in {{\mathbb {N}}}_{Lq}\}\) be the canonical orthonormal basis for \({{\mathbb {C}}}^{Lq}\). If \(P_{ker(\Psi )(x,\,\xi )}e_{n}=0\) for \(n\in {{\mathbb {N}}}_{Lq}\) and \((x,\,\xi )\in E\), then \(\textrm{rank}({\varvec{\Psi }})^*(x,\,\xi )=Lq\) for \((x,\,\xi )\in E\), this is a contradiction. Therefore, there exists \( n_0\in {{\mathbb {N}}}_{Lq}\) and \(E'\subset E\) with \(|E'|>0\) such that
Take \(c\in l^2(\{1,\,2,\ldots ,\,L\}\times {{\mathbb {Z}}^2})\) by \(C(x,\,\xi )=\left\{ \begin{array}{ll} P_{ker({\varvec{\Psi }}(x,\,\xi ))}e_{n_0} &{}\quad \hbox {if} \,\,(x,\xi )\in E'; \\ 0 &{}\quad \hbox {otherwise} \end{array} \right. \) for \((x,\,\xi )\in [1,\,b)\times [0,\,1)\). Then, \(C(x,\,\xi )\) is measurable on \([1,\,b)\times [0,\,1)\) by Lemma 2.8 and
Thus, \(C\in L^2([1,b)\times [0,1),\,{{\mathbb {C}}}^{Lq})\) and C is a nonzero solution to (2.24) in \(L^2([1,b)\times [0,1),\,{{\mathbb {C}}}^{Lq})\). The proof is completed. \(\square \)
By the same procedures as in ([24], Theorem 4.1) and Lemma 2.5, we have the following completeness and frame characterization of an \({{\mathcal{M}\mathcal{D}}}\)-system \({{\mathcal{M}\mathcal{D}}}(\Psi ,\,a,\,b)\).
Theorem 2.1
Let \(a,\,b\) and \(\Psi \) be as in the general setup. Then, (i) \({{\mathcal{M}\mathcal{D}}}(\Psi ,\,a,\,b)\) is complete in \(L^{2}({\mathbb {R}}_+)\) if and only if
(ii) \({{\mathcal{M}\mathcal{D}}}(\Psi ,\,a,\,b)\) is a Bessel sequence in \(L^{2}({\mathbb {R}}_+)\) with bound B if and only if
(iii) \({{\mathcal{M}\mathcal{D}}}(\Psi ,\,a,\,b)\) is a frame for \(L^{2}({\mathbb {R}}_+)\) with bounds A and B if and only if
Remark 2.3
(i) Observe that \({\varvec{\Psi }}(x,\,\xi )\) is a \(Lq\times p\) matrix. Theorem 2.1 (i) shows that the inequality
is necessary for the existence of complete \({\mathcal{M}\mathcal{D}}\)-systems in \(L^2({\mathbb {R}}_+)\). (ii) By Theorem 2.1 (iii), the lower frame bound A and upper frame bound B of a frame \({{\mathcal{M}\mathcal{D}}}(\Psi ,\,a,\,b)\) for \(L^2({\mathbb {R}}_+)\) always satisfy that
In particular, \({{\mathcal{M}\mathcal{D}}}(\Psi ,\,a,\,b)\) is a tight frame for \(L^2({\mathbb {R}}_+)\) only if \(q=1\).
Observe that a sequence is a Riesz basis if and only if it is a complete Riesz sequence, and if and only if it is a frame with Riesz property, and that an orthonormal basis (sequence) is exactly a Riesz basis (sequence) with bound 1. By Lemma 2.9 and Theorem 2.1, we have the following theorem.
Theorem 2.2
Let a, b and \(\Psi \) be as in the general setup. Suppose \({{\mathcal{M}\mathcal{D}}}(\Psi ,\,a,\,b)\) is a Bessel sequence in \(L^2({\mathbb {R}}_+)\). Then, the following are equivalent:
(i) \({{\mathcal{M}\mathcal{D}}}(\Psi ,\,a,\,b)\) is a Riesz basis (an orthonormal basis) for \(L^2({\mathbb {R}}_+)\).
(ii) \(\log _ba=L\), and \({{\mathcal{M}\mathcal{D}}}(\Psi ,\,a,\,b)\) is a Riesz sequence (an orthonormal sequence) in \(L^2({\mathbb {R}}_+)\).
(iii) \(\log _ba=L\), and \({{\mathcal{M}\mathcal{D}}}(\Psi ,\,a,\,b)\) is a frame (a Parseval frame) for \(L^2({\mathbb {R}}_+)\).
By Remark 2.3, if a and b are as in the general setup, \(p\le Lq\) is necessary for the existence of complete \({\mathcal{M}\mathcal{D}}\)-system of the form \({{\mathcal{M}\mathcal{D}}}(\Psi ,\,a,\,b)\). In view of Lemma 2.7, the following theorem provides a parametric expression of all complete \({\mathcal{M}\mathcal{D}}\)-systems and all \({\mathcal{M}\mathcal{D}}\)-frames in \(L^2({\mathbb {R}}_+)\) of the form \({{\mathcal{M}\mathcal{D}}}(\Psi ,\,a,\,b)\).
Theorem 2.3
Let \(a,\,b\) be as in the general setup, and \(\log _ba\le L\). Assume that \(\lambda _s(x,\,\xi )\in L^2([1,\,a^{\frac{1}{p}})\times [0,\,1))\) with \(s\in {\mathbb {N}}_{p}\). Define
and \(\Psi =\{\psi _1,\,\psi _2,\,\ldots ,\,\psi _L\}\) by an \(Lq\times p\) matrix-valued function
for \((x,\,\xi )\in [1,\,a^{\frac{1}{p}})\times [0,\,1)\), where \(U(x,\,\xi )\) and \(V(x,\,\xi )\) are \(Lq\times Lq\) and \(p\times p\) unitary matrix-valued measurable function defined on \([1,\,a^{\frac{1}{p}})\times [0,\,1)\), respectively. Then,
(i) \({{\mathcal{M}\mathcal{D}}}(\Psi ,\,a,\,b)\) is complete in \(L^{2}({\mathbb {R}}_+)\) if and only if
(ii) \({{\mathcal{M}\mathcal{D}}}(\Psi ,\,a,\,b)\) is a Bessel sequence in \(L^{2}({\mathbb {R}}_+)\) if and only if
In this case, the Bessel bound is
(iii) \({{\mathcal{M}\mathcal{D}}}(\Psi ,\,a,\,b)\) is a frame for \(L^{2}({\mathbb {R}}_+)\) if and only if
In this case, the frame bounds are
In particular, \({{\mathcal{M}\mathcal{D}}}(\Psi ,\,a,\,b)\) is a Riesz basis for \(L^{2}({\mathbb {R}}_+)\) if \(\log _{b}a=L\) in addition.
(iv) \({{\mathcal{M}\mathcal{D}}}(\Psi ,\,a,\,b)\) is a tight frame for \(L^{2}({\mathbb {R}}_+)\) if and only if \(q=1\) and there exists a constant \(A>0\) such that
In this case, the frame bound is A.
Proof
(i) and (iv) follow from Theorem 2.1. Next we prove (ii) and (iii). By a direct computation,
for a.e. \((x,\,\xi )\in [1,\,a^{\frac{1}{p}})\times [0,\,1).\) Observe that \(V(x,\,\xi )\) is unitary. It follows that
and
for a.e. \((x,\,\xi )\in [1,\,a^{\frac{1}{p}})\times [0,\,1).\) Again by Theorem 2.1, \({{\mathcal{M}\mathcal{D}}}(\Psi ,\,a,\,b)\) is a Bessel sequence in \(L^2({\mathbb {R}}_+)\) if and only if
and the Bessel bound is
in this case; \({{\mathcal{M}\mathcal{D}}}(\Psi ,\,a,\,b)\) is a frame for \(L^2({\mathbb {R}}_+)\) if and only if
and the frame bounds are
in this case. Next we prove
and
to finish the proof. Since
we have
On the other hand, since
we have
Hence, (2.29) holds. (2.30) may be proved similarly. And \({{\mathcal{M}\mathcal{D}}}(\Psi ,\,a,\,b)\) is a Riesz basis for \(L^{2}({\mathbb {R}}_+)\) if \(\log _{b}a=L\) in addition by Theorem 2.2. The proof is completed. \(\square \)
Collecting Theorem 2.3 and Remark 2.3, we obtain the following density theorem.
Theorem 2.4
Let \(a,\,b\) be as in the general setup. Then, the following are equivalent: (i) \(\log _ba\le L\). (ii) There exists a sequence \(\Psi =\{\psi _1,\,\psi _2,\,\ldots ,\,\psi _L\}\) in \(L^2({\mathbb {R}}_+)\) such that \({{\mathcal{M}\mathcal{D}}}(\Psi ,\,a,\,b)\) is complete in \(L^2({\mathbb {R}}_+)\). (iii) There exists a sequence \(\Psi =\{\psi _1,\,\psi _2,\,\ldots ,\,\psi _L\}\) in \(L^2({\mathbb {R}}_+)\) such that \({{\mathcal{M}\mathcal{D}}}(\Psi ,\,a,\,b)\) is a frame for \(L^2({\mathbb {R}}_+)\).
3 \({\mathcal{M}\mathcal{D}}\)-dual characterization
This section focuses on proving \(\log _ba\in {\mathbb {N}}\) being the necessary and sufficient condition for an \({\mathcal{M}\mathcal{D}}\)-frame to admit \({\mathcal{M}\mathcal{D}}\)-duals, and a parametric expression of \({\mathcal{M}\mathcal{D}}\)-duals. We first give some lemmas for later use. The first two represent the composition of the mixed frame operator \(S_{\Psi ,\,\Phi }\) and \(\Gamma \) (\(\Upsilon \), respectively) using \(\Theta _\beta \)-transform matrices.
Lemma 3.1
Let \(a,\,b\) and \(\Psi \) be as in the general setup, and \(\Phi =\{\phi _1,\,\phi _2,\,\ldots ,\,\phi _L\}\) be a sequence in \(L^2({\mathbb {R}}_+)\). Suppose \({{\mathcal{M}\mathcal{D}}}(\Psi ,\,a,\,b)\) and \({{\mathcal{M}\mathcal{D}}}(\Phi ,\,a,\,b)\) are both Bessel sequences in \(L^{2}({\mathbb {R}}_+)\). Then,
for \(f\in L^2({\mathbb {R}}_+)\), \(j\in {\mathbb {Z}}\) and a.e. \((x,\,\xi )\in [1,\,b)\times [0,\,1)\).
Proof
Fix an arbitrary \(f\in L^2({\mathbb {R}}_+)\). First we prove
Since \({{\mathcal{M}\mathcal{D}}}(\Psi ,\,a,\,b)\) and \({{\mathcal{M}\mathcal{D}}}(\Phi ,\,a,\,b)\) are both Bessel sequences in \(L^{2}({\mathbb {R}}_+)\), all entries of \({\varvec{\Psi }}(x,\,\xi ),\,\mathbf{\Phi }(x,\,\xi )\) belong to \(L^{\infty }({{\mathbb {R}}}_+\times {{\mathbb {R}}})\) by Remark 2.2. This implies that (2.18) holds for \(f,\,g\in L^2({{\mathbb {R}}}_+)\) by Lemma 2.2 (v). Thus, (2.17) holds, equivalently,
On the other hand,
by Lemma 2.2 (v). Thus, (3.2) holds.
Next we prove (3.1). Since \({{\mathbb {Z}}}={{\mathbb {N}}}_p+p{\mathbb {Z}}\), (3.1) is equivalent to, given \((s,\,j)\in {{\mathbb {N}}}_p\times {\mathbb {Z}}\) and \(f\in L^2({{\mathbb {R}}}_+)\),
for a.e. \((x,\,\xi )\in [1,\,b)\times [0,\,1)\). Now we prove (3.3). By Lemma 2.3 (iii), we have
and
for a.e. \((x,\,\xi )\in [1,\,b)\times [0,\,1)\). On the other hand,
for a.e. \((x,\,\xi )\in [1,\,b)\times [0,\,1)\) by Lemma 2.3 (iii) and (3.2). Substituting (3.4) and (3.5) into (3.6) gives (3.3). The proof is completed. \(\square \)
Lemma 3.2
Let \(a,\,b\) and \(\Psi \) be as in the general setup, \(\log _ba\in {\mathbb {N}}\) and \(\Phi =\{\phi _1,\,\phi _2,\,\ldots ,\,\phi _L\}\) be a sequence in \(L^2({\mathbb {R}}_+)\). Suppose \({{\mathcal{M}\mathcal{D}}}(\Psi ,\,a,\,b)\) and \({{\mathcal{M}\mathcal{D}}}(\Phi ,\,a,\,b)\) are both Bessel sequences in \(L^{2}({\mathbb {R}}_+)\). Then,
Proof
Suppose \(\log _ba\in {\mathbb {N}}\). Then, \(q=1\) and \({\mathbb {N}}_q=\{0\}\). And by Lemma 3.1, we have
for \((x,\,\xi )\in [1,\,a^{\frac{1}{p}})\times [0,\,1).\) The proof is completed. \(\square \)
The following lemma shows that \(\log _ba \in {\mathbb {N}}\) is the necessary and sufficient condition for dilation-and-modulation operation and frame operator to commute.
Lemma 3.3
Let \(a,\,b\) and \(\Psi \) be as in the general setup, and \(\psi _1,\,\psi _2,\ldots ,\psi _L\) be not all zero. Suppose \({{\mathcal{M}\mathcal{D}}}(\Psi , \,a,\,b)\) is a Bessel sequence in \(L^2({\mathbb {R}}_+)\). Then,
if and only if \(\log _ba \in {\mathbb {N}}\). In particular, if \({{\mathcal{M}\mathcal{D}}}(\Psi , \,a,\,b)\) is a frame for \(L^2({\mathbb {R}}_+)\), then
Proof
Sufficiency. Suppose \(\log _ba\in {{\mathbb {N}}}\). Then, \(a=b^p\) for some positive integer p. Since \(\Lambda _m\) is b-dilation periodic, for arbitrary \(m,\,j\in {\mathbb {Z}}\) and \(f\in L^2({{\mathbb {R}}}_+)\),
a.e. on \({{\mathbb {R}}}_+\). Thus,
Arbitrarily fix \(m_0\), \(j_0\in {\mathbb {Z}}\) and \(f\in L^2({\mathbb {R}}_+)\). By (3.9), we have
This leads to (3.8) by the arbitrariness of \(m_0\), \(j_0\) and f.
Necessity. By contradiction. Suppose (3.8) holds, but \(\log _{b}a\notin {\mathbb {N}}\). Then, \(q>1\) and \(1\in {{\mathbb {N}}}_q\). Arbitrarily fix \(m,\,j\in {\mathbb {Z}}\) and \(f\in L^2({{\mathbb {R}}}_+)\). By Lemmas 3.1, 2.2 (iv) and 2.3 (i), we have
for a.e. \((x,\,\xi )\in [1,\,b)\times [0,\,1)\). On the other hand, by Lemma 2.2 (iv),
for a.e. \((x,\,\xi )\in [1,\,b)\times [0,\,1)\). By (3.8), we have
This implies that
by (3.10) and (3.11). Since \(\psi _1,\,\psi _2,\,\ldots ,\psi _L\) are not all zero, \({\varvec{\Psi }}(x,\,\xi )\) is a nonzero matrix-valued function on \([1,\,a^{\frac{1}{p}})\times [0,\,1)\) by Remark 1.1, equivalently, there exists \(E_0\subset [1,\,a^{\frac{1}{p}})\times [0,\,1)\) with \(|E_0|>0\) such that
Let \(E=\left\{ (ax,\,\xi ):\,(x,\,\xi )\in E_0\right\} \). Thus, \(E\subset [a,\,a^{\frac{1}{p}+1})\times [0,\,1)\) with \(|E|>0\) satisfies
by (3.13) and Lemma 2.3 (i). Since
is a partition of \({\mathbb {R}}_+\), there exist \(k\in {\mathbb {Z}}\) and \(s\in {{\mathbb {N}}}_p\) such that
So, by Lemma 3.1,
for a.e. \((ax,\,\xi )\in E\cap \left( [b^{kp+s},\,b^{kp+s+1})\times [0,\,1)\right) \). This implies that
for a.e. \((ax,\,\xi )\in E\cap ([b^{kp+s},\,b^{kp+s+1})\times [0,\,1)) \) by (3.12). Since (3.14) holds, there exists \( s_0\in {{\mathbb {N}}}_p\) and \(E_1\subset E\cap \left( [b^{kp+s},\,b^{kp+s+1})\times [0,\,1)\right) \) with \(|E_1|>0\) such that the \(s_0\)th column \(\left( \overline{{\varvec{\Psi }^*}(ax,\,\xi ){\varvec{\Psi }}(ax,\,\xi )}\right) _{s_0}\) of \(\overline{{\varvec{\Psi }^*}(ax,\,\xi ){\varvec{\Psi }}(ax,\,\xi )}\) satisfies
Let \(\{e_k:\,k\in {\mathbb {N}}_p\}\) be the canonical orthonormal basis for \({{\mathbb {C}}}^p\). Take f in (3.16) by
Thus, \(f\in L^2({\mathbb {R}}_+)\) by Lemma 2.2 (v), and
And then, by (3.16),
This implies that
by (3.17), equivalently,
Thus, \(\log _ba\in {{\mathbb {N}}}\) due to \(\log _ba>0\). It is a contradiction. The other part follows from the fact that \(S_{\Psi }\) is invertible if \({{\mathcal{M}\mathcal{D}}}(\Psi ,\,a,\,b)\) is a frame for \(L^2({{\mathbb {R}}}_+)\). The proof is completed. \(\square \)
The following theorem gives a characterization of the existence of \({\mathcal{M}\mathcal{D}}\)-duals for a given \({\mathcal{M}\mathcal{D}}\)-frame.
Theorem 3.1
Let \(a,\,b\) and \(\Psi \) be as in the general setup. Suppose \({{\mathcal{M}\mathcal{D}}}(\Psi ,\,a,\,b)\) is a frame for \(L^2({\mathbb {R}}_+)\). Then, there exists a sequence \(\Phi =\{\phi _1,\,\phi _2,\,\ldots ,\,\phi _L\}\) in \(L^2({\mathbb {R}}_+)\) such that \({{\mathcal{M}\mathcal{D}}}(\Phi ,\,a,\,b)\) is a dual of \({{\mathcal{M}\mathcal{D}}}(\Psi ,\,a,\,b)\) if and only if \(\log _ba\in {{\mathbb {N}}}\).
Proof
Sufficiency. Suppose \(\log _ba\in {{\mathbb {N}}}\). Then,
by Lemma 3.3. Take \(\Phi =\{\phi _1,\,\phi _2,\,\ldots ,\,\phi _L\}\) as \(\phi _l=S_{\Psi }^{-1}\psi _l\) for \(1\le l\le L\). Then, \({{\mathcal{M}\mathcal{D}}}({\Phi },\,a,\,b)\) is a dual of \({{\mathcal{M}\mathcal{D}}}(\Psi ,\,a,\,b)\).
Necessity. By contradiction. Suppose \(\log _ba\notin {{\mathbb {N}}}\) and \({{\mathcal{M}\mathcal{D}}}(\Phi ,\,a,\,b)\) is a dual of \({{\mathcal{M}\mathcal{D}}}(\Psi ,\,a,\,b)\). Then, \([a^{\frac{1}{p}},\,a^{\frac{2}{p}})\subset [1,\,b)\) due to \(b=a^{\frac{p}{q}}\), and
for a.e. \((x,\,\xi )\in [1,\,b)\times [0,\,1)\) by Lemma 3.1. Let \(\{e_k:\,k\in {{\mathbb {N}}}_p\}\) is the canonical orthonormal basis for \({{\mathbb {C}}}^p\). Fix an arbitrary \(k\in {{\mathbb {N}}}_p\). Take f in (3.18) by
Then, it is well defined due to \(\Gamma \) being a unitary operator from \(L^2({\mathbb {R}}_+)\) onto \(L^2([1,\,b)\times [0,\,1),\,{{\mathbb {C}}}^p)\) by Lemma 2.2 (v), and (3.18), the kth column of \({\varvec{\Psi }}^{*}(x,\,\xi ){\varvec{\Phi }}(x,\,\xi )\) is \(e_k\). Thus,
By Lemma 2.3 (ii), this implies that
for a.e. \((x,\,\xi )\in [1, \,a^{\frac{1}{p}})\times [0,\,1)\), equivalently,
This contradicts (3.19) by the fact that \(b=a^{\frac{q}{p}}\) and \([a^{\frac{1}{p}},\,a^{\frac{2}{p}})\subset [1,\,b)\). The proof is completed. \(\square \)
Remark 3.1
By the proof of Theorem 3.1, if \(\log _{b}a\in {\mathbb {N}}\) (this implies \(b=a^{\frac{1}{p}}\)), two Bessel sequences \({{\mathcal{M}\mathcal{D}}}(\Psi ,\,a,\,b)\) and \({{\mathcal{M}\mathcal{D}}}(\Phi ,\,a,\,b)\) in \(L^2({\mathbb {R}}_+)\) form a pair of dual frames for \(L^2({\mathbb {R}}_+)\) if and only if
By Theorems 2.4 and 3.1, under the general setup, \(\log _ba=p\le L\) is the sufficient and necessary condition for the existence of \({\mathcal{M}\mathcal{D}}\)-duals of a given \({\mathcal{M}\mathcal{D}}\)-frame. The following theorem presents a parametric expression of all \({\mathcal{M}\mathcal{D}}\)-duals of an arbitrarily given \({\mathcal{M}\mathcal{D}}\)-frame and shows that the window functions of the canonical dual have minimal norm among all \({\mathcal{M}\mathcal{D}}\)-duals.
Theorem 3.2
Let \(a,\,b\) and \(\Psi \) be as in the general setup, \(\log _{b}a=p\le L\), and \({{\mathcal{M}\mathcal{D}}}(\Psi , \,a,\,b)\) be a frame for \(L^2({\mathbb {R}}_+)\). Then,
(i) All \({\mathcal{M}\mathcal{D}}\)-dual frames \({{\mathcal{M}\mathcal{D}}}(\Phi ,\,a,\,b)\) of \({{\mathcal{M}\mathcal{D}}}(\Psi ,\,a,\,b)\) are precisely defined by
for a.e. \((x,\,\xi )\in [1,\,a^{\frac{1}{p}})\times [0,1)\), where \({\varvec{\Omega }}(x,\,\xi )\) is an \(Lq\times p\) matrix-valued function with all entries in \(L^{\infty }([1,\,a^{\frac{1}{p}})\times [0,\,1))\). In particular, \({\varvec{\Psi }}(x,\,\xi )({\varvec{\Psi }}^*(x,\,\xi ){\varvec{\Psi }}(x,\,\xi ))^{-1}\) is exactly the \(\Theta _{\beta }\)-transform matrix associated with \(S_\Psi ^{-1}\Psi \), which is exactly the window function of the canonical dual \({{\mathcal{M}\mathcal{D}}}(S_\Psi ^{-1}\Psi ,\,a,\,b)\).
(ii) For each \({\mathcal{M}\mathcal{D}}\)-dual frame \({{\mathcal{M}\mathcal{D}}}(\Phi ,\,a,\,b)\) of \({{\mathcal{M}\mathcal{D}}}(\Psi ,\,a,\,b)\),
Proof
(i) Since \({{\mathcal{M}\mathcal{D}}}(\Psi ,\,a,\,b)\) is a frame for \(L^2({\mathbb {R}}_+)\),
for some \(0<A\le B<\infty \) by Theorem 2.1 and the fact that \(q=1\). First we prove that each \(\Phi =\{\phi _1,\,\phi _2,\,\ldots ,\,\phi _L\}\) defined by (3.21) generates an \({\mathcal{M}\mathcal{D}}\)-dual \({{\mathcal{M}\mathcal{D}}}(\Phi ,\,a,\,b)\) of \({{\mathcal{M}\mathcal{D}}}(\Psi ,\,a,\,b)\). By (3.23) and Remark 2.2, all entries of \({\varvec{\Phi }}(x,\,\xi )\) defined by (3.21) belong to \(L^{\infty }([1,\,a^{\frac{1}{p}})\times [0,\,1))\). Thus, \({{\mathcal{M}\mathcal{D}}}(\Phi ,\,a,\,b)\) is a Bessel sequence in \(L^{2}({\mathbb {R}}_+)\). A direct computation gives
This implies that \({{\mathcal{M}\mathcal{D}}}(\Phi ,\,a,\,b)\) is a dual frame of \({{\mathcal{M}\mathcal{D}}}(\Psi ,\,a,\,b)\). Next we prove that the window function \(\Phi =\{\phi _1,\,\phi _2,\,\ldots ,\,\phi _L\}\) of each \({{\mathcal{M}\mathcal{D}}}\)-dual frame \({{\mathcal{M}\mathcal{D}}}(\Phi ,\,a,\,b)\) of \({{\mathcal{M}\mathcal{D}}}(\Psi ,\,a,\,b)\) may be defined by (3.21). Suppose \({{\mathcal{M}\mathcal{D}}}(\Phi ,\,a,\,b)\) is a dual frame of \({{\mathcal{M}\mathcal{D}}}(\Psi ,\,a,\,b)\). Then,
by Remark 3.1. Take \({\varvec{\Omega }}(x,\,\xi )={\varvec{\Phi }}(x,\,\xi )\). Then, all entries of \({\varvec{\Omega }}(x,\,\xi )\) belong to \(L^{\infty }([1,\,a^{\frac{1}{p}})\times [0,\,1))\) by Remark 2.2, and \({\varvec{\Phi }}(x,\,\xi )\) has the form (3.21) by (3.24) and a direct computation.
In particular, by Lemma 3.2, we have
It follows that
Thus,
for a.e. \((x,\,\xi )\in [1,\,a^{\frac{1}{p}})\times [0,\,1)\).
(ii) Suppose \({{\mathcal{M}\mathcal{D}}}(\Phi ,\,a,\,b)\) is a dual frame of \({{\mathcal{M}\mathcal{D}}}(\Psi ,\,a,\,b)\). By Lemma 3.3, the canonical dual frame of \({{\mathcal{M}\mathcal{D}}}(\Psi ,\,a,\,b)\) is \({{\mathcal{M}\mathcal{D}}}(S_\Psi ^{-1}\Psi ,\,a,\,b)\). This implies that
for \(1\le l\le L\). Observe that
by Lemma 3.3. It follows that
and thus
for \(1\le l\le L\). This implies that
due to \({{\mathcal{M}\mathcal{D}}}(\Phi ,\,a,\,b)\) being a dual frame of \({{\mathcal{M}\mathcal{D}}}(\Psi ,\,a,\,b)\). If follows that
This leads to (3.22). The proof is completed. \(\square \)
Recall from the Gabor analysis theory that an arbitrary Gabor frame for \(L^2({\mathbb {R}})\) has Gabor duals ([3], Theorem 12.3.2). Theorem 3.1 shows that this cannot be completely extended to \({\mathcal{M}\mathcal{D}}\)-frames for \(L^2({\mathbb {R}}_+)\), and it tells us that, under the general setup, an \({\mathcal{M}\mathcal{D}}\)-frame has \({\mathcal{M}\mathcal{D}}\)-duals if and only if \(\log _ba\in {\mathbb {N}}\). Therefore, \({\mathcal{M}\mathcal{D}}\)-frames in \(L^2({\mathbb {R}}_+)\) are essentially different from Gabor frames in \(L^2({\mathbb {R}})\). Recall that a Gabor frame for \(L^2({\mathbb {R}})\) is a Riesz basis if and only if it has the unique Gabor dual. The following theorem shows that an \({\mathcal{M}\mathcal{D}}\)-frame for \(L^2({\mathbb {R}}_+)\) is a Riesz basis if and only if it admits the unique \({\mathcal{M}\mathcal{D}}\)-dual. It presents that Gabor frames and \({\mathcal{M}\mathcal{D}}\)-frames share some similarity.
Theorem 3.3
Let \(a,\,b\) and \(\Psi \) be as in the general setup, \(\log _{b}a=p\le L\), and \({{\mathcal{M}\mathcal{D}}}(\Psi , \,a,\,b)\) be a frame for \(L^2({\mathbb {R}}_+)\). Then, the following are equivalent:
(i) \({{\mathcal{M}\mathcal{D}}}(\Psi ,\,a,\,b)\) is a Riesz basis for \(L^2({\mathbb {R}}_+)\).
(ii) \({{\mathcal{M}\mathcal{D}}}(\Psi ,\,a,\,b)\) has the unique \({\mathcal{M}\mathcal{D}}\)-dual \({{\mathcal{M}\mathcal{D}}}(S_{\Psi }^{-1}\Psi ,\,a,\,b)\).
(iii) \(\log _ba=L\).
Proof
By Theorem 2.2, (i) and (iii) are equivalent. Next we prove the equivalence between (ii) and (iii). First we prove that (ii) implies (iii). Suppose (ii) holds. By Theorem 3.2, for each \(L\times p\) matrix-valued function \({\varvec{\Omega }}(x,\,\xi )\) with \(L^{\infty }([1,\,a^{\frac{1}{p}})\times [0,\,1))\)-entries,
Let \(\{e_k:\,k\in {{\mathbb {N}}}_L\}\) be the canonical orthonormal basis for \({{\mathbb {C}}}^L\). Fix an arbitrary \(k\in {{\mathbb {N}}}_L\), take \({\varvec{\Omega }}(x,\,\xi )\) in (3.25) by
Then, the kth column of \({\varvec{\Psi }}(x,\,\xi )\left( {\varvec{\Psi }}^*(x,\,\xi ){\varvec{\Psi }}(x,\,\xi )\right) ^{-1}{\varvec{\Psi }}^*(x,\,\xi )\) is exactly \(e_k\). This implies that
Thus,
for a.e. \((x,\,\xi )\in [1,\,a^{\frac{1}{p}})\times [0,\,1)\) by Theorem 2.1. This leads to (iii) by the fact that \(\log _ba=p\le L\).
Next, we prove that (iii) implies (ii). Suppose (iii) holds. Then, \({\varvec{\Psi }}(x,\,\xi )\) is an \(L\times L\) matrix-valued function. Since \(\left( {\varvec{\Psi }}^*(x,\,\xi ){\varvec{\Psi }}(x,\,\xi )\right) ^{-1}\) exists for a.e. \((x,\,\xi )\in [1,\,a^{\frac{1}{p}})\times [0,\,1)\) by Theorem 2.1 (iii), so do \({\varvec{\Psi }}(x,\,\xi )\) and \({\varvec{\Psi }}^*(x,\,\xi )\). It follows that
for a.e. \((x,\,\xi )\in [1,\,a^{\frac{1}{p}})\times [0,\,1)\). Thus, for all \({\varvec{\Omega }}(x,\,\xi )\) with \(L^{\infty }([1,\,a^{\frac{1}{p}})\times [0,\,1))\)-entries,
for a.e. \((x,\,\xi )\in [1,\,a^{\frac{1}{p}})\times [0,\,1)\). Therefore, \({{\mathcal{M}\mathcal{D}}}(\Psi ,\,a,\,b)\) admits the unique \({{\mathcal{M}\mathcal{D}}}\)-dual \({{\mathcal{M}\mathcal{D}}}(S_{\Psi }^{-1}\Psi ,\,a,\,b)\) by Theorem 3.2. The proof is completed. \(\square \)
Theorems 2.4 and 3.1 show that, under the general setup, \(\log _ba=p\le L\) is necessary for the existence of \({\mathcal{M}\mathcal{D}}\)-dual frames. Theorem 3.2 reduces constructing \({\mathcal{M}\mathcal{D}}\)-dual of a given \({\mathcal{M}\mathcal{D}}\)-frame \({{\mathcal{M}\mathcal{D}}}(\Psi ,\,a,\,b)\) to computing \({\varvec{\Psi }}(x,\,\xi )({\varvec{\Psi }}^*(x,\,\xi ){\varvec{\Psi }}(x,\,\xi ))^{-1}\) for \((x,\,\xi )\in [1,\,a^{\frac{1}{p}})\times [0,\,1)\) which corresponds to the canonical dual. An alternate \({\mathcal{M}\mathcal{D}}\)-dual may be obtained by choosing \({\varvec{\Omega }}(x,\,\xi )\) with \(L^{\infty }([1,\,a^{\frac{1}{p}})\times [0,\,1))\)-entries. The following theorem gives the expression of the canonical \({\mathcal{M}\mathcal{D}}\)-dual of a given \({\mathcal{M}\mathcal{D}}\)-frame. By Lemma 2.7, it covers all situations.
Theorem 3.4
Let a, b as be in the general setup with \(\log _ba=p\le L\). Assume \(\lambda _s(x,\,\xi )\) with \(s\in {\mathbb {N}}_{p}\) are measurable functions satisfying
Define
\(\Psi =\{\psi _1,\,\psi _2,\,\ldots ,\,\psi _L\}\) by a \(L\times p\) matrix-valued function
for \((x,\,\xi )\in [1,\,a^{\frac{1}{p}})\times [0,1)\), where \(U(x,\,\xi )\) and \(V(x,\,\xi )\) are \(L\times L\) and \(p\times p\) unitary matrix-valued measurable function defined on \([1,\,a^{\frac{1}{p}})\times [0,1)\), respectively. Then, \({{\mathcal{M}\mathcal{D}}}(\Psi ,\,a,\,b)\) is a frame for \(L^2({\mathbb {R}}_+)\), and the window functions of its canonical dual \({{\mathcal{M}\mathcal{D}}}(S_\Psi ^{-1}\Psi ,\,a,\,b)\) are defined by the \(\Theta _\beta \)-transform matrix
where \({\widetilde{\Lambda }}(x,\xi )=\textrm{diag}\left( \overline{\lambda _0^{-1}(x,\,\xi )},\,\overline{\lambda _1^{-1}(x,\,\xi )},\,\cdots ,\, \overline{\lambda _{p-1}^{-1}(x,\,\xi )}\right) .\)
Proof
By Theorem 2.3, \({{\mathcal{M}\mathcal{D}}}(\Psi ,\,a,\,b)\) is a frame for \(L^2({\mathbb {R}}_+)\). By (3.27), we have
and thus
for a.e. \((x,\,\xi )\in [1,\,a^{\frac{1}{p}})\times [0,\,1)\). This leads to the theorem by Theorem 3.2. The proof is completed. \(\square \)
In what follows, we give some examples of Theorem 3.4.
Example 3.1
Given a, \(b>1\) satisfying \(\log _ba=3\), define \(\lambda _0(x,\,\xi )=(1+x)e^{2\pi i\xi }\), \(\lambda _1(x,\,\xi )=2x\), \(\lambda _2(x,\,\xi )=(a-x)e^{-2\pi i\xi }\) for \((x,\,\xi )\in [1,\,a^{\frac{1}{3}})\times [0,\,1)\) in Theorem 3.4.
(i) Take
for a.e. \((x,\,\xi )\in [1,\,a^{\frac{1}{3}})\times [0,\,1)\). It is clearly that \(U(x,\,\xi )\) and \(V(x,\,\xi )\) are unitary matrix-valued measurable function defined on \([1,\,a^{\frac{1}{3}})\times [0,1)\). Define \(\Psi =\{\psi _1,\,\psi _2,\,\psi _3\}\) by (3.27). Then,
for a.e. \((x,\,\xi )\in [1,\,a^{\frac{1}{3}})\times [0,\,1)\), \({{\mathcal{M}\mathcal{D}}}(\Psi , \,a,\,b)\) is a Riesz basis for \(L^2({\mathbb {R}}_+)\) by Theorem 2.2, and by Remark 1.1 and a standard computation, \(\Psi =\{\psi _1,\,\psi _2,\,\psi _3\}\) has the form
and
By Theorem 3.4, the window functions \(S_{\Psi }^{-1}\Psi =\left\{ S_{\Psi }^{-1}\psi _1,\,S_{\Psi }^{-1}\psi _2,\,S_{\Psi }^{-1}\psi _3\right\} \) of its canonical dual may be defined by the \(\Theta _{\beta }\)-transform
for \((x,\,\xi )\in [1,\,a^{\frac{1}{3}})\times [0,\,1)\). Applying Remark 1.1, we have
(ii) Take \(U(x,\,\xi )=V(x,\,\xi )=I_3\) for \((x,\,\xi )\in [1,\,a^{\frac{1}{3}})\times [0,\,1)\), and define \(\Psi =\{\psi _1,\,\psi _2,\,\psi _3\}\). Then, \({{\mathcal{M}\mathcal{D}}}(\Phi ,\,a,\,b)\) is a Riesz basis for \(L^2({\mathbb {R}}_+)\) by Theorem 2.2. Similarly to (i), we have
and the window functions \(S_{\Psi }^{-1}\Psi =\left\{ S_{\Psi }^{-1}\psi _1,\,S_{\Psi }^{-1}\psi _2,\,S_{\Psi }^{-1}\psi _3\right\} \) have the form
Example 3.2
Given \(a,\,b>1\) satisfying \(\log _ba=2\), define
for a.e. \((x,\,\xi )\in [1,\,a^{\frac{1}{2}})\times [0,\,1)\) in Theorem 3.4. Clearly, \(U(x,\,\xi )\) and \(V(x,\,\xi )\) are unitary.
(i) Take \(\lambda _0(x,\,\xi )=e^{2\pi i\xi }\), \(\lambda _1(x,\,\xi )=1\) for a.e. \((x,\,\xi )\in [1,\,a^{\frac{1}{2}})\times [0,\,1)\). Define \(\Psi =\{\psi _l:\,1\le l\le L\}\) by (3.27). Then,
for a.e. \((x,\,\xi )\in [1,\,a^{\frac{1}{2}})\times [0,\,1)\), \({{\mathcal{M}\mathcal{D}}}(\Psi ,\,a,\,b)\) is a Parseval frame for \(L^2({\mathbb {R}}_+)\) by Theorem 2.3 (v). And by Remark 1.1 and a standard computation,
(ii) Take \(\lambda _0(x,\,\xi )=(1+x)e^{2\pi i\xi }\), \(\lambda _1(x,\,\xi )=2x\) for a.e. \((x,\,\xi )\in [1,\,a^{\frac{1}{2}})\times [0,\,1)\), and define \(\Psi =\{\psi _1,\,\psi _2,\,\psi _3\}\) by (3.27). Then, \({{\mathcal{M}\mathcal{D}}}(\Psi ,\,a,\,b)\) is a frame for \(L^2({\mathbb {R}}_+)\) by Theorem 3.4. Similarly to (i), we have
and by Theorem 3.4 the window functions \(S_{\Psi }^{-1}\Psi =\left\{ S_{\Psi }^{-1}\psi _1,\,S_{\Psi }^{-1}\psi _2,\,S_{\Psi }^{-1}\psi _3\right\} \) have the form
Availability of data and materials
Not applicable.
References
Barbieri, D., Hern\(\acute{\text{a}}\)ndez, E., Paternostro, V.: The Zak transform and the structure of spaces invariant by the action of an LCA group. J. Funct. Anal. 269, 1327–1358 (2015)
Bownik, M.: A characterizations of affine dual frames in \(L^2({{\mathbb{R} }}^n)\). Appl. Comput. Harmon. Anal. 8, 203–221 (2000)
Christensen, O.: An Introduction to Frames and Riesz Bases, Second edition, Birkh\(\ddot{\text{ a }}\)user, (2016)
Christensen, O.: Pairs of dual Gabor frames with compact support and desired frequency localization. Appl. Comput. Harmon. Anal. 20, 403–410 (2006)
Christensen, O., Kim, R.Y.: On dual Gabor frame pairs generated by polynomials. J. Fourier Anal. Appl. 16, 1–16 (2010)
Christensen, O., Kim, H.O., Kim, R.Y.: On entire functions restricted to intervals, partition of unities, and dual Gabor frames. Appl. Comput. Harmon. Anal. 38, 72–86 (2015)
Chui, C.K., Czaja, W., Maggioni, M., Weiss, G.: Characterization of general tight wavelet frames with matrix dilations and tightness preserving oversampling. J. Fourier Anal. Appl. 8, 173–200 (2002)
Daubechies, I.: The wavelet transform, time-frequency localization and signal analysis. IEEE Trans. Inform. Theory 36, 961–1005 (1990)
Daubechies, I., Han, B.: The canonical dual frame of a wavelet frame. Appl. Comput. Harmon. Anal. 12, 269–285 (2002)
Daubechies, I., Han, B.: Pairs of dual wavelet frames from any two refinable functions. Constr. Approx. 20, 325–352 (2004)
Daubechies, I., Landau, H.J., Landau, Z.: Gabor time-frequency lattices and the Wexler-Raz identity. J. Fourier Anal. Appl. 1, 437–478 (1995)
Duffin, R.J., Schaeffer, A.C.: A class of nonharmonic Fourier series. Trans. Amer. Math. Soc. 72, 341–366 (1952)
Frazier, M., Garrig\(\acute{\text{ o }}\)s, G., Wang, K., Weiss, G.: A characterization of functions that generate wavelet and related expansion. J. Fourier Anal. Appl. 3, 883–906 (1997)
Gertner, I., Tolimiieri, R.: Multiplicative Zak transform. J. Vis. Commun. Image R 6, 89–95 (1995)
Han, B.: Homogeneous wavelets and framelets with refinable structure. Sci. China Math. 60, 2173–2189 (2017)
Heil, C.: A basis theory primer, Expanded edition, Birkh\({\ddot{\text{ a }}}\)user, New York, (2011)
Hern\(\acute{\text{ a }}\)ndez, E., Labate, D., Weiss, G.: A unified characterization of reproducing systems generated by a finite family. II. J. Geom. Anal. 12, 615–662 (2002)
Hern\(\acute{\text{ a }}\)ndez, E., \(\breve{\text{ S }}\)iki\(\acute{\text{ c }}\), H., Weiss, G., Wilson, E.: Cyclic subspaces for unitary representations of LCA groups; generalized Zak transform. Colloq. Math. 118, 313–332 (2010)
Janssen, A.J.E.M.: Duality and biorthogonality for Weyl-Heisenberg frames. J. Fourier Anal. Appl. 1, 403–436 (1995)
Janssen, A.J.E.M.: Signal analytic proofs of two basic results on lattice expansions. Appl. Comput. Harmon. Anal. 1, 350–354 (1994)
Labate, D.: A unified characterization of reproducing systems generated by a finite family. J. Geom. Anal. 12, 469–491 (2002)
Li, S.: On general frame decompositions. Numer. Funct. Anal. Optim. 16, 1181–1191 (1995)
Li, Y.-Z., Li, Y.-N.: Dilation-and-modulation frame sets on the half real line. Math. Methods Appl. Sci. 45, 6998–7023 (2022)
Li, Y.-Z., Wang, Y.-H.: The density theorem of a class of dilation-and-modulation systems on the half real line. Results Math. 4, 19 pp (2019)
Li, Y.-Z., Wang, Y.-H.: The dilation-and-modulation frame sets on the half real line. Acta Math. Sin. (Chin. Ser.) 63, 16 pp (2020)
Li, Y.-Z., Zhang, W.: Dilation-and-modulation systems on the half real line. J. Inequal. Appl. 186, 1–11 (2016)
Li, Y.-Z., Zhang, W.: Multi-window dilation-and-modulation frames on the half real line. Sci China Math. 62, 1–16 (2019)
Rudin, W.: Real and complex analysis, 3rd edn. McGraw-Hill Book Co., New York (1987)
Stoeva, D.T.: On compactly supported dual windows of Gabor frames. J. Math. Anal. Appl. 505, 10 pp (2022)
Wang, Y.-H., Li, Y.-Z.: A class of vector-valued dilation-and-modulation frames on the half real line. Math. Methods Appl. Sci. 41, 3900–3912 (2018)
Wexler, J., Raz, S.: Discrete Gabor expansions. Signal Process 21, 207–220 (1990)
Young, R.M.: An Introduction to Nonharmonic Fourier Series. Academic Press, New York (1980)
Funding
This work was supported by the National Natural Science Foundation of China (Grant No. 11971043).
Author information
Authors and Affiliations
Contributions
All authors contributed equally to the writing of this paper. All authors read and approved the final manuscript.
Corresponding author
Ethics declarations
Conflicts of interest
The authors declare that they have no conflicts of interest.
Code availability
Not applicable.
Additional information
Communicated by Rosihan M. Ali.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Supported by National Natural Science Foundation of China (Grant No. 11971043).
Rights and permissions
Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
About this article
Cite this article
Li, YN., Li, YZ. A class of reproducing systems generated by a finite family in \(L^2({\mathbb {R}}_+)\). Bull. Malays. Math. Sci. Soc. 46, 99 (2023). https://doi.org/10.1007/s40840-023-01493-3
Received:
Revised:
Accepted:
Published:
DOI: https://doi.org/10.1007/s40840-023-01493-3
Keywords
- Frame
- Riesz basis
- \({\mathcal{M}\mathcal{D}}-system\)
- \({\mathcal{M}\mathcal{D}}\)-frame
- \({\mathcal{M}\mathcal{D}}\)-dual frame