2025-11-15T12:49:10.652717
Rational Orthonormal Littlewood-Paley Wavelet Basis of L2(R)
Sun
In this letter, first, we prove that the orthonormal basis of rational Littlewood-Paley wavelet with rational dilation factor M=p/q first proposed by Auscher does not hold for all rational numbers. It does not hold if q is not equal to 1. In other words, it is not an orthonormal basis if the rational dilation factor M is not an integer. Then, to make up for the shortcoming of the rational Littlewood-Paley wavelet proposed by Auscher, a new orthonormal basis of rational Littlewood-Paley wavelet with rational dilation factor M=p/q is proposed, which holds for all rational numbers. Finally, by means of sampling theorem for bandpass signals, it is proved completely that the new rational Littlewood-Paley wavelet family is an orthonormal wavelet basis of L2(R).
academicRational Orthonormal Littlewood-Paley Wavelet Basis of L2(R)
Basic Information
- Paper ID: 2511.08984
- Title: Rational Orthonormal Littlewood-Paley Wavelet Basis of L2(R)
- Author: Sun Changping (Department of Artificial Intelligence, School of Electrical Engineering and Automation, Tianjin University of Technology)
- Classification: math.FA (Functional Analysis)
- Publication Time/Conference: Not explicitly marked; submitted as IEEE member contribution
- Paper Link: https://arxiv.org/abs/2511.08984
Abstract
This paper first proves that the rational Littlewood-Paley wavelet orthonormal basis with rational dilation factor M=p/q proposed by Auscher does not apply to all rational numbers; when q≠1, this basis does not hold. In other words, when the rational dilation factor M is not an integer, it is not an orthonormal basis. To remedy the defects in Auscher's rational Littlewood-Paley wavelet, this paper proposes a new rational Littlewood-Paley wavelet orthonormal basis with rational dilation factor M=p/q that applies to all rational numbers. Finally, through the bandpass signal sampling theorem, the paper completely proves that the new rational Littlewood-Paley wavelet family is an orthogonal wavelet basis of L2(R).
Research Background and Motivation
1. Core Problem to be Solved
This paper aims to address the construction of orthonormal bases for rational Littlewood-Paley wavelets in rational multiresolution analysis. Specifically, it seeks to correct the mathematical defects in the rational Littlewood-Paley wavelet basis first proposed by Auscher in his 1989 doctoral dissertation.
2. Importance of the Problem
- Theoretical Significance: Rational multiresolution analysis compared to traditional dyadic analysis provides better signal component separation capability
- Practical Applications: Rational wavelet transforms have been applied to fault diagnosis, automatic imagined speech recognition, automatic detection of cardiac valve diseases, and other fields
- Algorithm Validation: Orthonormal rational Littlewood-Paley wavelet basis is a key tool for verifying the validity of rational wavelet decomposition and reconstruction algorithms
3. Limitations of Existing Methods
The rational Littlewood-Paley wavelet basis proposed by Auscher has fatal defects:
- When q≠1 in the rational dilation factor M=p/q, the basis does not satisfy orthonormality
- It is an orthonormal basis only when q=1 (i.e., M is an integer)
- This causes signal decomposition and reconstruction using this basis to produce erroneous results when the dilation factor is non-integer
4. Research Motivation
Although rational wavelet research is extensive, the rational Littlewood-Paley wavelets cited in the literature remain in Auscher's form, lacking rigorous orthonormality proofs. The author discovered this theoretical gap and proposes a correction scheme to ensure theoretical completeness and application correctness.
Core Contributions
- Proof of Theoretical Defect: Rigorously proves that Auscher's rational Littlewood-Paley wavelet basis does not satisfy orthonormality when q≠1 (Theorem 1)
- New Wavelet Basis Construction: Proposes a new rational Littlewood-Paley wavelet orthonormal basis with rational dilation factor M=p/q applicable to all rational numbers, providing explicit expressions in both frequency and time domains
- Completeness Proof: Using the bandpass signal sampling theorem, completely proves that the newly proposed rational Littlewood-Paley wavelet family is an orthogonal wavelet basis of L2(R) (Theorem 2), including:
- Orthogonality proof
- Normality proof
- Completeness proof
- Theoretical Unification: Proves that Auscher's wavelet basis is a special case of the proposed wavelet basis when q=1 (Corollary), achieving theoretical unification
Detailed Methodology
Task Definition
Construct a rational Littlewood-Paley wavelet family with rational dilation factor M=p/q (where p,q are coprime positive integers):
such that it satisfies:
- Orthogonality: when
- Normality:
- Completeness: spans the entire L2(R) space
Model Architecture
Frequency Domain Representation
The new rational Littlewood-Paley wavelet is defined in the frequency domain as:
\sqrt{q}, & \text{if } |\omega| \in [M, M^2) \\ 0, & \text{otherwise} \end{cases}$$ where the key parameters are: - $\omega_1 = M = p/q$ (lower frequency bound) - $\omega_2 = M^2 = p^2/q^2$ (upper frequency bound) #### Time Domain Representation According to the inverse Fourier transform, the time-domain wavelet function is: $$\psi(t) = \sqrt{q} \cdot \frac{\sin(\omega_2 t) - \sin(\omega_1 t)}{\pi t}$$ $$= \sqrt{q} \cdot \frac{\sin(M^2 t) - \sin(Mt)}{\pi t}$$ #### Wavelet Family Generation Wavelet basis functions are generated through dilation and translation: $$\psi_{j,k}(t) = M^{j/2} \psi(M^j t - k)$$ Its frequency domain representation is: $$\hat{\psi}_{j,k}(\omega) = M^{-j/2} e^{-i\omega k/M^j} \hat{\psi}(\omega/M^j)$$ Support set: $\text{supp}(\hat{\psi}_{j,k}) = [M^{j+1}, M^{j+2}) \cup [-M^{j+2}, -M^{j+1})$ ### Technical Innovation Points #### 1. Correction of Normalization Factor **Key Innovation**: Introduction of the $\sqrt{q}$ factor in the frequency domain representation - **Auscher's Form**: $\hat{\psi}(\omega) = \mathbb{1}_{[M, M^2)}(\omega)$ (indicator function) - **New Form**: $\hat{\psi}(\omega) = \sqrt{q} \cdot \mathbb{1}_{[M, M^2)}(\omega)$ **Reason for Correction**: By computing the L2 norm: $$\|\psi\|_{L^2}^2 = \frac{1}{2\pi}\int_{-\infty}^{\infty}|\hat{\psi}(\omega)|^2 d\omega = \frac{1}{2\pi} \cdot 2 \int_M^{M^2} q \, d\omega = \frac{q}{\pi}(M^2 - M)$$ When the $\sqrt{q}$ factor is introduced, $\|\psi\|_{L^2} = 1$ can be satisfied. #### 2. Orthogonality Proof Strategy Employs frequency domain analysis combined with support set separation: - **Case A**: When $j \neq j'$, utilizes disjointness of support sets - **Case B**: When $j = j', k \neq k'$, utilizes translation invariance - **Case C**: Comprehensive analysis utilizing Parseval's formula #### 3. Innovation in Completeness Proof **Core Idea**: Transforms the wavelet decomposition problem into a bandpass signal sampling problem - Fix scale j and define subspace $V_j$ as the set of functions supported on $[M^{j+1}, M^{j+2}]$ - Identify functions in $V_j$ as bandpass signals with bandwidth $B = M^{j+2} - M^{j+1}$ - Apply the bandpass signal sampling theorem to determine sampling frequency $f_s = \frac{2(M^{j+2} - M^{j+1})}{n_s}$ - Select appropriate $n_s$ such that the sampling period matches the wavelet translation parameter - Prove that $\{\psi_{j,k}\}_{k \in \mathbb{Z}}$ forms an orthonormal basis of $V_j$ - Since $\bigcup_{j=-\infty}^{\infty} V_j = L^2(\mathbb{R})$, completeness is established ## Experimental Setup ### Theoretical Verification Framework This paper is a pure mathematical theory paper employing rigorous mathematical proofs rather than experimental verification. Main verification methods include: 1. **Counterexample Construction**: Proves that Auscher's wavelet L2 norm is not 1 when q≠1 through computation 2. **Direct Calculation**: Computes inner products and norms of the new wavelet basis 3. **Theoretical Derivation**: Utilizes classical theories such as Parseval's formula and sampling theorem ### Visualization Display The paper provides visualization of two concrete examples (Figures 1 and 2): - **Example 1**: M=3/2, m=1,2 - **Example 2**: M=5/3, m=1,2 Demonstrating the characteristics of the new wavelet in the frequency domain. ## Experimental Results ### Main Results (Theoretical Proofs) #### Theorem 1 (Defect of Auscher's Wavelet) **Conclusion**: The rational Littlewood-Paley wavelet basis proposed by Auscher is not an orthonormal basis when q≠1. **Proof Highlights**: Computing yields $\|\psi\|_{L^2}^2 = \frac{1}{\pi}(M^2 - M) = \frac{1}{\pi} \cdot \frac{p(p-q)}{q^2}$ When q≠1, $\|\psi\|_{L^2} \neq 1$, violating the normality condition. #### Theorem 2 (Orthonormality of New Wavelet Basis) **Conclusion**: The newly proposed rational Littlewood-Paley wavelet family $\{\psi_{j,k}\}_{j,k \in \mathbb{Z}}$ is an orthogonal wavelet basis of L2(R). **Proof Structure**: 1. **Orthogonality Proof** (Formulas 7-8): - For all $(j,k) \neq (j',k')$, proves $\langle \psi_{j,k}, \psi_{j',k'} \rangle = 0$ - Discusses three cases: $j \neq j'$, $j = j'$ and $k \neq k'$, and comprehensive cases - Utilizes disjointness of frequency domain support sets 2. **Normality Proof** (Formula 9): $$\|\psi_{j,k}\|_{L^2}^2 = \frac{1}{2\pi} \int_{-\infty}^{\infty} |\hat{\psi}_{j,k}(\omega)|^2 d\omega = 1$$ 3. **Completeness Proof** (Formulas 10-11): - Introduces subspace $V_j = \{f \in L^2(\mathbb{R}) : \text{supp}(\hat{f}) \subseteq [M^{j+1}, M^{j+2}]\}$ - Utilizes the bandpass signal sampling theorem, selecting sampling frequency $\omega_s = 2(M^{j+2} - M^{j+1})$ - Sampling period $T = \frac{2\pi}{\omega_s} = \frac{1}{M^j}$ - Proves that $\{\psi_{j,k}\}_{k \in \mathbb{Z}}$ is an orthonormal basis of $V_j$ - Obtains completeness from $\bigcup_{j=-\infty}^{\infty} V_j = L^2(\mathbb{R})$ #### Corollary (Theoretical Unification) **Conclusion**: Auscher's rational Littlewood-Paley wavelet basis is orthonormal if and only if q=1. **Proof**: Comparing formulas (1) and (3), they are identical when q=1; combining Theorems 1 and 2 yields the corollary. ### Case Analysis The paper demonstrates the properties of the new wavelet through two concrete examples: **Example 1**: M=3/2 (p=3, q=2) - Frequency domain support: $[3/2, 9/4) \cup [-9/4, -3/2)$ - Normalization factor: $\sqrt{2}$ - Figure 1 shows the wavelet shape for m=1,2 **Example 2**: M=5/3 (p=5, q=3) - Frequency domain support: $[5/3, 25/9) \cup [-25/9, -5/3)$ - Normalization factor: $\sqrt{3}$ - Figure 2 shows the wavelet shape for m=1,2 ### Experimental Findings 1. **Critical Role of Normalization Factor**: The $\sqrt{q}$ factor is crucial for ensuring orthonormality 2. **Applicability of Bandpass Signal Theory**: Transforming the wavelet analysis problem into a bandpass signal sampling problem is an effective proof strategy 3. **Universality of Theory**: The new wavelet basis applies to all rational dilation factors M=p/q without the restriction q=1 ## Related Work ### 1. Foundations of Rational Multiresolution Analysis - **Auscher (1989, 1992)**: First proposed the concept of rational multiresolution analysis and rational Littlewood-Paley wavelets, but with theoretical defects - **Baussard et al. (2004)**: Proposed pyramid algorithms for rational multiresolution analysis and fast wavelet transforms, applied to wavelet shrinkage denoising ### 2. Completeness Theory of Orthogonal Wavelet Systems - **Laugesen (2001)**: Provided characterization conditions for completeness of orthogonal wavelet systems with arbitrary real dilation factors - **Chui & Shi (2000)**: Completely characterized tight frames and orthogonal wavelets with arbitrary real dilation factors - **Li (2014)**: Provided perfect reconstruction conditions for orthogonal wavelets with rational dilation factors based on MRA ### 3. Rational Wavelet Transform Algorithms - **Bayram & Selesnick (2009)**: Developed overcomplete discrete wavelet transforms with rational dilation factors - **Li et al. (2008)**: Proposed fast algorithms for discrete wavelet transforms with dilation factor 3/2, overcoming frequency distortion in high-frequency subbands of Mallat DWT ### 4. Application Domains - **Fault Diagnosis** (Sangeetha, 2019): Three-phase induction motor fault diagnosis based on rational dilation wavelet transforms - **Speech Recognition** (Kamble et al., 2023): Optimized rational dilation wavelet transforms for automatic imagined speech recognition - **Medical Diagnosis** (Zeng et al., 2023): Automatic detection of cardiac valve diseases using rational dilation wavelet transforms ### Relationship of This Work to Related Work - **Corrects Foundational Theory**: Rectifies theoretical defects in Auscher's pioneering work - **Provides Rigorous Proofs**: First complete proof that rational Littlewood-Paley wavelet family is an orthogonal basis of L2(R) - **Completes Theory**: Ensures mathematical rigor of rational wavelet theory, providing reliable theoretical foundation for applications ## Conclusions and Discussion ### Main Conclusions 1. **Clarification of Theoretical Defect**: Auscher's rational Littlewood-Paley wavelet basis is orthonormal only when q=1 (i.e., M is an integer), not applicable to general rational dilation factors 2. **Construction of New Basis**: The proposed new rational Littlewood-Paley wavelet basis, through introduction of the $\sqrt{q}$ normalization factor, applies to all rational numbers M=p/q 3. **Complete Proof**: Using the bandpass signal sampling theorem, first completely proves that the rational Littlewood-Paley wavelet family is an orthogonal wavelet basis of L2(R) 4. **Theoretical Unification**: Auscher's wavelet basis is a special case of the new wavelet basis when q=1, achieving theoretical unification ### Limitations 1. **Limited to Littlewood-Paley Type**: This paper addresses only Littlewood-Paley-type wavelets (rectangular window in frequency domain), not other types of rational wavelets 2. **Lack of Numerical Experiments**: As a pure theoretical paper, lacks numerical experiments verifying the performance of the new wavelet basis in practical signal processing 3. **Fast Algorithms Not Provided**: Although a new wavelet basis is proposed, corresponding fast decomposition and reconstruction algorithms are not provided 4. **Insufficient Application Verification**: Does not demonstrate advantages of the new wavelet basis over Auscher's wavelet in specific applications (such as denoising, compression, etc.) ### Future Directions The paper explicitly indicates future research directions: 1. **Fast Algorithm Research**: Based on prior work [8], investigate analysis and synthesis fast algorithms for the new rational wavelet transform 2. **Algorithm Verification**: Use the new rational Littlewood-Paley wavelet to verify the validity of new fast algorithms 3. **Application Extension**: Apply the new wavelet basis to practical signal processing tasks and evaluate its performance ## In-Depth Evaluation ### Strengths #### 1. Significant Theoretical Contribution - **Discovery of Important Defect**: Identifies fundamental mathematical errors in the widely-cited Auscher wavelet basis in the literature - **Provides Complete Correction**: Not only identifies the problem but provides a corrected scheme applicable to all rational numbers - **Rigorous Proofs**: Provides complete mathematical proofs of orthogonality, normality, and completeness #### 2. Innovation in Proof Methods - **Interdisciplinary Fusion**: Cleverly introduces bandpass signal sampling theorem into wavelet analysis, providing new insights for completeness proofs - **Clear Structure**: Proofs are hierarchical and case-by-case, with rigorous logic #### 3. Theoretical Unification - Demonstrates through corollary that Auscher's wavelet is a special case of the new wavelet, achieving elegant theoretical unification - Maintains compatibility with classical integer dilation wavelet theory #### 4. Practical Significance - Provides correct theoretical foundation for applications of rational wavelet transforms - Prevents algorithm verification failures caused by using incorrect wavelet bases ### Shortcomings #### 1. Missing Experimental Verification - **Insufficient Numerical Verification**: Lacks numerical computation verifying orthonormality of the new wavelet basis - **Missing Performance Comparison**: Does not compare performance of new wavelet basis versus Auscher's wavelet in practical applications - **Limited Visualization**: Provides only frequency domain plots for two examples, lacking time-domain waveforms and scale-translation family visualizations #### 2. Expandable Theoretical Depth - **Limited to L2(R)**: Does not discuss properties in other function spaces (such as Sobolev spaces) - **Missing Regularity Analysis**: Does not analyze smoothness, vanishing moments, and other important properties of the new wavelet - **Frame Theory Not Addressed**: Does not discuss construction of corresponding tight frames or redundant dictionaries #### 3. Gaps in Algorithm Level - **Missing Fast Algorithms**: Does not provide fast decomposition and reconstruction algorithms for the new wavelet basis - **Computational Complexity Not Analyzed**: Does not discuss computational efficiency of the new wavelet basis - **Implementation Details Insufficient**: Lacks guidance for practical programming implementation #### 4. Improvable Writing - **Symbol Usage**: Some symbol definitions are not sufficiently clear (such as first appearance of ω₁, ω₂ in formula (5)) - **Figure Quality**: Figures 1 and 2 have low resolution and insufficient annotations - **Insufficient Comparison**: Does not provide intuitive comparison plots between Auscher's and new wavelets ### Impact #### 1. Contribution to the Field - **Correction of Theoretical Foundation**: Rectifies fundamental errors in rational wavelet theory spanning over 30 years, with far-reaching impact - **Guidance for Subsequent Research**: Provides correct starting point for theoretical research and algorithm design of rational wavelets - **Application Reliability**: Ensures applications based on rational wavelets (fault diagnosis, speech recognition, etc.) have solid theoretical support #### 2. Practical Value - **Algorithm Verification Tool**: Can be used to verify correctness of rational wavelet decomposition and reconstruction algorithms - **Benchmark Testing**: Can serve as benchmark for evaluating other rational wavelet construction methods - **Teaching Value**: Can serve as typical case in wavelet theory teaching #### 3. Reproducibility - **Theory Verifiable**: Mathematical proofs are clear and easy to verify - **Formulas Explicit**: Frequency and time domain expressions are explicit, easy to implement programmatically - **Parameters Concrete**: Provides concrete examples (M=3/2, 5/3), facilitating reproduction #### 4. Potential Limitations - **Application Scope**: Applies only to Littlewood-Paley-type wavelets; other types of rational wavelets require separate investigation - **Computational Efficiency**: Lacks fast algorithms; practical applications may be limited by computational efficiency - **Generalization Difficulty**: Extension to high dimensions or other transforms requires additional theoretical work ### Applicable Scenarios #### 1. Theoretical Research - Rational multiresolution analysis theory research - Wavelet basis construction method research - Frame theory and sampling theory research #### 2. Algorithm Development - Theoretical verification of rational wavelet transform algorithms - Design of new rational wavelet decomposition and reconstruction algorithms - Correctness verification of fast algorithms #### 3. Signal Processing Applications - Signal analysis requiring non-integer scale decomposition - Application scenarios requiring flexible frequency resolution - Multi-scale signal representation and feature extraction #### 4. Specific Domain Applications - **Fault Diagnosis**: Mechanical equipment vibration signal analysis - **Biomedical**: Physiological signal processing (ECG, EEG, etc.) - **Speech Processing**: Speech recognition, speech enhancement - **Image Processing**: Texture analysis, image denoising #### 5. Inapplicable Scenarios - Applications requiring compactly supported wavelets (Littlewood-Paley wavelets lack compact support in time domain) - Real-time applications with extreme computational efficiency requirements (lacking fast algorithms) - Applications requiring high-order vanishing moments (vanishing moment properties not analyzed) ## References ### Key References 1. **Auscher, P. (1989)**: "Ondelettes fractales et applications", Ph.D. Thesis - Pioneering work in rational wavelet theory 2. **Auscher, P. (1992)**: "Wavelet bases for L2(R) with rational dilation factor" - Original literature corrected in this paper 3. **Baussard et al. (2004)**: "Rational multiresolution analysis and fast wavelet transform" - Important work on fast algorithms for rational wavelets 4. **Laugesen (2001)**: "Completeness of orthonormal wavelet systems for arbitrary real dilations" - Theoretical foundation for completeness 5. **Proakis & Manolakis (2006)**: "Digital signal processing" - Reference source for bandpass signal sampling theorem --- ## Summary This paper is an important theoretical correction work that rectifies a fundamental error in rational wavelet theory that has existed for over 30 years. By introducing a simple yet critical normalization factor $\sqrt{q}$, the author constructs a Littlewood-Paley wavelet orthonormal basis applicable to all rational dilation factors and provides complete mathematical proofs using the bandpass signal sampling theorem. The paper's primary value lies in theoretical rigor and correctness, providing a reliable foundation for subsequent research and applications of rational wavelets. Although lacking experimental verification and fast algorithms, as a pure theoretical work, this paper's contribution is significant and necessary. Future work should focus on algorithm implementation, performance evaluation, and practical application verification.