Abstract
In this paper, we begin with the classical concept of tight frames in Hilbert spaces. First, we introduce the orthogonal projection P between H and (the range of the frame transform associated with a traditional tight frame) and investigate the relationship between P and . We then explore fusion frames and extend the index set to an infinite set through a concrete example. Second, we examine the role of orthogonal projections in fusion frames with particular emphasis on robustness and redundancy illustrated by examples. Finally, we study dual fusion frames and establish several important results, especially concerning the relationship between the frame operators of two types of dual fusion frames.
MSC:
46H25; 42C15
1. Introduction
Frame theory plays an important role in signal processing, image processing, data compression, sampling theory, and related fields. In 1952, Duffin and Schaeffer [1] introduced the concept of frames in Hilbert spaces in their study of non-harmonic Fourier series. However, this idea initially received little attention. It was not until 1986 that Daubechies, Grossman, and Meyer [2] carried out groundbreaking work in this area, after which frame theory began to be extensively studied.
Casazza and Kutyniok [3] later introduced fusion frames and investigated their properties, perturbations, and the approximation of inverse operators. In recent years, the theory of fusion frames has also undergone rapid development [4,5,6,7]. Fusion frames generalize traditional frames and are particularly well suited for processing block data or multi-channel signals, as they combine orthogonal projections of subspaces with weights.
More recently, considerable attention has been devoted to the study of dual frames. For example, J. Lopez, Han Deguang, Sun Wenchang, and others [8,9,10,11,12,13,14,15,16] have investigated the problem of selecting optimal dual frames from various perspectives. Their contributions have enabled frame theory to achieve significant progress in practical applications.
Orthogonal projection also plays a central role in frame theory. It is fundamental in contexts ranging from rigorous decomposition in Hilbert spaces to efficient approximation in machine learning, and from time-frequency localization in multiresolution analysis to geometric dimensionality reduction in optimization problems. Orthogonal projection provides an indispensable theoretical foundation for the modeling, analysis, and control of complex systems. By balancing redundancy and efficiency through orthogonality constraints, it continues to play a key role in fields such as signal processing, communications, and control [17].
Throughout this paper, H is a Hilbert spaces, we write as in Section 3 for the convenience, and denote by . The symbol denotes the identity operator on H. We now present the preliminary definitions and theorems used in this paper.
Definition 1
(see [18]). A sequence for a Hilbert space H is called a frame if there exist constants such that for any ,
The constants a and b are called the lower and the upper frame bounds, respectively. A frame is called tight frame if and a normalized tight frame if .
Definition 2
(see [18]). Let be a frame for a Hilbert space H, and let be its orthonormal basis. The associated frame transform θ is defined by
Then θ is adjointable, its adjoint is given by
where .
Moreover, we have
A direct calculation now yields
If is a frame for a Hilbert space H, then the operator is a positive, self-adjoint and invertible operator on H, called the frame operator.
It follows from Definition 2 that is a-tight frame if and only if and a normalized tight frame if and only if .
Since , using to replace f, we obtain
and
Then,
We now recall the definition of the dual frame.
Definition 3
(see [18]). Let be a frame for a Hilbert space H with the frame operator S. Then is called the canonical dual frame of . If a frame for H satisfies
then is called an alternate dual frame of .
Definition 4
(see [19]). Let I be a countable index set, let be a family of closed subspaces of H, and let be a family of weights. Then is a fusion frame if there exist constants so that
where is the orthogonal projection. The constants A and B are called the lower and the upper fusion frame bounds, respectively.
The family is called A-tight fusion frame if and a parseval fusion frame if . If only the right-hand inequality holds, it is called a Bessel fusion sequence with Bessel frame bound B. Family is called 1-consistent Parseval fusion frame if
or w-consistent Parseval fusion frame if
Theorem 1
(see [20]). Let H be a Hilbert space, and let be orthogonal projections. If for any , . Then is a frame for H if and only if is fusion frame for H with the frame bounds unchanged and the weight set being .
Theorem 2
(see [20]). Let be a sequence of closed subspaces of H, be the weight set, and let be frames for . Then is a fusion frame for H if and only if is a frame for H.
2. Results
2.1. Discovery of Fusion Frames on Hilbert Spaces and Generalization of the Index Set
Theorem 3.
Let be a-tight frame for H with the frame transform θ, and let be an orthogonal projection. Then and , where is the orthonormal basis for H.
Proof.
Since is a-tight frame for H, and on , i.e., . Then for any ,
therefore, or .
Since , we obtain
By the arbitrariness of f, it follows that .
In addition, from and , we have
Its self-adjointness is obvious. Thus, , which fully verifies that P acts as an orthogonal projection in the relevant context. □
Let be -tight frame sequence for H with the frame transforms (where i is the number of frames and j is the dimension of the frames), and let be the orthogonal projections. Then , and
Therefore, is a fusion frame if and only if there exist constants so that
Obviously, since , when I is a finite set, must be a Bessel fusion sequence.
Our questions are, do such traditional frames and fusion frames exist? And, can I here be extended to an infinite set, i.e., Our answers are affirmative and we now give example below.
Example 1.
Let with and , Then is a s-tight fusion frame for H, where is the standard orthonormal basis for H.
Proof.
For any , we have
so are -tight frames for H. Since
i.e., , . Then
which means that satisfying and
Therefore, is a s-tight fusion frame.
However, cannot be 1-uniform fusion frame or w-uniform fusion frame, because the constant series diverges forever unless I is a finite set. □
We present it in the form of a certain theorem.
Theorem 4.
Let H be a Hilbert space. Then there exists a sequence of -tight frames for H such that is a tight fusion frame for H, where are the frame transforms of , are the orthogonal projections.
Remark 1.
(1) Example 1 makes use of the convergence of positive-term series. For instance, if we take , then , and forms a Parseval fusion frame, and in this case, .
In fact, for the positive term-series considered here, it is sufficient to establish convergence, there is no need to compute the exact sum. Moreover, many series can only be analyzed in terms of convergence, without an explicit evaluation of their sums. Even in such cases, still constitutes a (tight) fusion frame.
(2) The conclusion only holds for traditional tight frames, since the relationship between the frame transform and orthogonal projection exists exclusively in the tight-frame setting.
(3) Example 1 and Theorem 4 establish a close connection between traditional frames and fusion frames, and further realize the closed subspaces of H by taking . Even more significantly, the index set I is generalized to infinite sets.
2.2. Orthogonal Projections in the Fusion Frames
The orthogonal projector P is a self-adjoint operator satisfying , i.e., . It projects element of H onto a closed subspace, and the projection is orthogonal to the complement of that subspace, that is, the inner product of the residual with the subspace is zero. Below, we present some simple examples in and .
Example 2.
In , let . Then for any and for any ,
Example 3.
In , let . Then
for any where the coefficients are the Fourier coefficients.
Example 4.
In , let be a fixed interval, and let
Then
A fusion frame is a subspace frame. Its central role lies in providing stable signal representation and reconstruction through subspace decomposition and weight balancing, while preserving the characteristics of each data source. Fusion frames thus serve as an important theoretical tool in multimodal and multi-sensor data processing. The following examples illustrate the role of orthogonal projection in the fusion frames and in the projection process as a whole.
Example 5.
Let , , (i.e., is the x-axis and is the y-axis), Then for any , , , , . Thus, the weighted sum is
Therefore, is a Parseval fusion frame.
If fails, and only and are retained, then the weighted sum is , which cannot control . However, if a third subspace is added (i.e., is the line ()). Let , then and the weighted sum is
Since
Therefore, is a fusion frame, and the lower and upper fusion frame bounds are , , respectively. In other words, when fails, f can still be reconstructed through and , which demonstrates the robustness of the fusion frame.
In addition, we can also add two subspaces , (i.e., is the line , is the line ), Then,
and the weighted sum is
Thus,
Therefore, is a fusion frame. In other words, f can still be reconstructed through , , and .
Orthogonal projection is also a fundamental tool for constructing tight frames. For instance, redundant frames can be transformed into Parseval frames by means of orthogonal projection. This property ensures lossless reconstruction in signal processing.
Example 6.
Let , , , , . Then for any , the weighted sum is
Therefore, is a fusion frame, but not a Parseval fusion frame.
Since can be linearly represented by and , it is therefore a redundant frame. Next, we transform it into a Parseval frame. we derive the analysis operator T of f. Since
Thus, T is a symmetric matrix, and
Projecting onto the original subspaces yields
After adjusting the weights , the weighted sum is obtained as
Thus, by projecting the subspaces of the redundant frame via and adjusting the weights, the redundant frame is converted into a Parseval fusion frame.
2.3. Research on Dual Fusion Frames on Hilbert Space
In this section, we begin with an example to explore how traditional dual frames in Hilbert spaces recover original signals in cases of data loss demonstrating the importance and research significance of dual frames.
Example 7.
Let , , . Then
so is a tight frame for with frame bound , where is the frame transform of X and S is its frame operator.
Let the original signal vector be . Then the frame coefficients are , , . Suppose that during transmission, the original signal loses , how can we recover the original signal?
We try to recover it using the alternate dual frame of X.
Since is lost, let us set , , . Then . Since Y is the alternate dual frame of X, they are mutually alternate dual frames and or . Then we have
It is solved that , , , . So , , , and
i.e., the alternate dual frame Y of X can recover the original signal.
The canonical dual frame of X can also recover the original signal.
Since , . Then we obtain
Taking
then
This means that the canonical dual frame Z of X can also recover the original signal.
Meanwhile, let (). Then
and
We obtain , i.e., , and of course , where .
It follows that the frames X and U are disjointed, meaning that any alternate dual frame of a frame can be expressed as the sum of its canonical dual frame and a frame can disjoint from itself, and the canonical dual frame is the “minimal” dual frame.
Next, we study dual fusion frames on Hilbert spaces.
Definition 5
(see [19]). Let be a fusion frame for H, the analysis operator is defined by
where . It can easily be shown that the synthesis operator , which is defined to be the adjoint operator, is given by
The fusion frame operator S for is defined by
If be a fusion frame for H with fusion frame bound A and B, then the associated fusion frame operator S is a self-adjoint, positive, and invertible operator on H, and .
It follows from Definition 5 that is a B-tight fusion frame if and only if , and is a Parseval fusion frame if and only if .
Obviously, if is a fusion frame for H with fusion frame operator S, then is a Parseval fusion frame for H. In fact,
Theorem 5
(see [1]). Let T be a bounded linear operator on H, and let V be a closed subspace of H. Then . Moreover if T is a unitary operator, then , where P is the orthogonal projection from H to V.
It follows from Theorem 5 that , i.e., . So, if is the fusion frame with fusion frame operator S, then
Therefore, for any , , and
is called the canonical dual fusion frame of .
Corollary 1.
Let be the fusion frame for H with fusion frame operator S, and let be its the canonical dual fusion frame. Then .
Proof.
For any ,
Therefore, the frame operator of the canonical dual fusion frame is the inverse of its own frame operator. It is obvious that if is a parseval fusion frame, i.e., , then its canonical dual fusion frame is itself. □
Definition 6
(see [16]). Let and be fusion frames for H. If for any ,
then is called an alternate dual fusion frame of , where S is the fusion frame operator of .
Remark 2.
(1) The canonical dual fusion frame of must be its alternate dual fusion frame. In particular, when and , these two dual fusion frames are the same fusion frame.
(2) Traditional alternate dual frames are mutual alternate duals, while alternate dual fusion frames are not dual to each other.
In fact, if is the alternate dual fusion frame of , then for any , where is the fusion frame operator of . If is an alternate dual fusion frame of , then for any , where is the fusion frame operator of .
In particular, if , then the alternate dual fusion frames are mutual alternate duals.
The following are the necessary and sufficient conditions under which they are mutual alternate duals and their stability.
Theorem 6.
Let and be fusion frames for H, let and be their analysis operators respectively, let and be their synthesis operators, respectively, and let and be their frame operators respectively. Then
(1) and are mutual alternate duals if and only if .
(2) If and is alternate dual fusion frames of , then is also a fusion frame for H.
Proof.
(1) For any , we have
Therefore, and are mutual alternate duals if and only if
if and only if
by the arbitrariness of f, we obtain .
(2) Since , and are mutual alternate duals. Combining with (1) of Theorem 6, for any , we have
And
Therefore,
that is to say,
Moreover, since and are fusion frames for H, there exist constants such that . Therefore,
and
Similarly, since
we have
so is also a fusion frame with frame operator and . □
Theorem 7.
Let be a fusion frame for H with fusion frame operator , let and be its canonical dual fusion frame and alternate dual fusion frame respectively. Then , where is the fusion frame operator of .
Proof.
Since for any ,
and
we have
Replacing f with ,
Therefore,
and
that is to say,
Moreover, since
we obtain
Thus,
Furthermore, we have
And
therefore,
Since is the frame operator of the canonical dual fusion frame of , Theorem 7 reveals the relationship between the canonical dual fusion frame and the alternate dual fusion frame, especially the relationship between their respective fusion frame operators. □
3. Conclusions
In this paper, we first uncover implicit fusion frames by studying the frame transform and orthogonal projection of traditional tight frames. Next, we present an example connecting traditional tight frames and fusion frames by combining the convergence of positive-term series in mathematical analysis with the orthonormal basis of a Hilbert space. In this context, the index set is extended to an infinite set and both types of frames are explicitly realized. Secondly, we investigated orthogonal projections in fusion frames, particularly their applications in transforming redundant fusion frames into Parseval fusion frames. Finally, we studied the dual fusion frames. Since alternate dual fusion frames are not dual to each other, we first established the necessary and sufficient conditions under which they are mutual alternate duals, and then examined their stability as well as the relationship between the two types of dual fusion frames.
Author Contributions
Writing—original draft, F.D.; Writing—review and editing, F.D.; formal analysis, R.P. All authors have read and agreed to the published version of the manuscript.
Funding
This work is supported by the National Natural Science Foundation of China (No. 11661070).
Data Availability Statement
The original contributions presented in this study are included in the article. Further inquiries can be directed to the corresponding author.
Acknowledgments
We thank two anonymous referees for their comments to improve the manuscript.
Conflicts of Interest
The authors declare no conflicts of interest.
References
- Duffin, R.J.; Schaeffer, A.C. A class of nonharmonic Fourier series. Trans. Am. Math. Soc. 1952, 72, 341–366. [Google Scholar] [CrossRef]
- Daubechies, I.; Grossmann, A.; Meyer, Y. Painless nonorthogonal expansions. J. Math. Phys. 1986, 27, 1271–1283. [Google Scholar] [CrossRef]
- Casazza, P.G.; Kutyniok, G. Frames of subspaces. Contemp. Math. 2004, 345, 87–114. [Google Scholar]
- Casazza, P.G.; Kutyniok, G.; Rozell, C.J. Modeling Sensor Networks with Fusion Frames. In Proceedings of the SPIE: International Society for Optics and Photonics, San Diego, CA, USA, 26–30 August 2007; SPIE: Bellingham, WA, USA, 2007; pp. 216–221. [Google Scholar]
- Casazza, P.G.; Kutyniok, G. Robustness of fusion frames under erasures of subspaces and of local frame vectors. Contemp. Math. 2008, 464, 149–160. [Google Scholar][Green Version]
- Casazza, P.G.; Fickus, M. Minimizing fusion frame potential. Acta Appl. Math. 2009, 107, 7–24. [Google Scholar] [CrossRef]
- Cahill, J.; Casazza, P.G.; Li, S. Non-orthogonal fusion frames and the sparsity of fusion frame operators. J. Fourier Anal. Appl. 2012, 18, 287–308. [Google Scholar] [CrossRef][Green Version]
- Lopez, J.; Han, D. Optimal dual frames for erasures. Linear Algebra Its Appl. 2010, 432, 471–482. [Google Scholar] [CrossRef]
- Han, D.; Larson, D.R. Frames, Bases and Group Representations; American Mathematical Society: Providence, RI, USA, 2000. [Google Scholar]
- Han, D.; Sun, W. Reconstruction of signals from frame coefficients with erasures at unknown locations. IEEE Trans. Inf. Theory 2014, 60, 4013–4025. [Google Scholar] [CrossRef]
- Leng, J.; Han, D.; Huang, T. Optimal dual frames for communication coding with probabilistic erasures. IEEE Trans. Signal Process. 2011, 59, 5380–5389. [Google Scholar] [CrossRef]
- Leng, J.; Han, D.; Huang, T. Probability modelled optimal frames for erasures. Linear Algebra Its Appl. 2013, 438, 4222–4236. [Google Scholar] [CrossRef]
- Sargazi, M.; Hossein, J. Fattahi Abdolmajid. Fusion Frame Duals and Its Applications to Invertible Bessel Fusion Multipliers. Iran. J. Sci. Technol. 2021, 45, 945–954. [Google Scholar] [CrossRef]
- Shamsabadi, M.; Arefijamaal, A.A.; Sadeghi, G. Duals and multipliers of K-fusion frames. J. Pseudo-Differ. Oper. Appl. 2020, 11, 1621–1633. [Google Scholar] [CrossRef]
- Arabyani-Neyshaburi, F.; Arefijamaal, A.A. A New Look at Optimal Dual Problem Related to Fusion Frames. Real Anal. Exch. 2024, 49, 87–110. [Google Scholar] [CrossRef]
- Ghosh, P.; Samanta, T.K. Stability of dual controlled g-fusion frames in hilbert spaces. Turk. World Math. Soc. J. Appl. Eng. Math. 2024, 14, 154–172. [Google Scholar]
- Fraser, J.M. Applications of dimension interpolation to orthogonal projections. Res. Math. Sci. 2025, 12, 10. [Google Scholar] [CrossRef] [PubMed]
- Christensen, O. An Introduction to Frames and Riesz Bases; Birkhauser: Cham, Switzerland, 2008. [Google Scholar]
- Casazza, P.G.; Kutyniok, G.; Li, S. Fusion frames and distributed processing. Appl. Comput. Harmon. Anal. 2008, 25, 114–132. [Google Scholar] [CrossRef]
- Gavruta, P. On the duality of fusion frames. J. Math. Anal. Appl. 2007, 333, 871–879. [Google Scholar] [CrossRef]
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content. |
© 2025 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/).