Wavelet multipliers and signals

He Zhao, M. W. WongThe Journal of the Australian Mathematical Society Series B Applied Mathematics 1999 · DOI

Abstract: The Schatten-von Neumann property of a pseudo-differential operator is established by showing that the pseudo-differential operator is a multiplier defined by means of an admissible wavelet associated to a unitary representation of the additive group $\mathbb{R}^n$ on the $C^*$-algebra of all bounded linear operators from $L^2(\mathbb{R}^n)$ into $L^2(\mathbb{R}^n)$. A bounded linear operator on $L^2(\mathbb{R})$ arising in the Landau, Pollak and Slepian model in signal analysis is shown to be a wavelet multiplier studied in this paper.

Introduction

In their 1999 paper, Z. He and M. W. Wong provide a rigorous mathematical bridge between abstract operator theory and the practical requirements of signal processing. By framing pseudo-differential operators as "wavelet multipliers," the authors demonstrate that these operators possess desirable spectral properties. Crucially, they show that a cornerstone of signal analysis—the Landau-Pollak-Slepian operator—is a specific instance of this broader theoretical framework.

The Problem: Operator Compactness and Signal Constraints

In both functional analysis and signal processing, operators are used to transform functions and signals. A significant challenge lies in determining the "size" or "compactness" of these operators. If an operator belongs to a Schatten-von Neumann class ($S_p$), it behaves similarly to a finite-dimensional matrix, which simplifies the study of its eigenvalues and eigenvectors.

He and Wong specifically investigate pseudo-differential operators, which generalize differential operators used to model physical systems. Their goal is to define these operators using wavelets and determine the conditions under which they exhibit the Schatten-von Neumann property. Furthermore, they seek to unify this theory with the Landau-Pollak-Slepian model, the standard framework for analyzing signals restricted in both time and frequency domains.

Methodological Framework

The authors employ several sophisticated tools from functional analysis to establish their results:

  • Unitary Representations: The operators are defined using a unitary representation of the additive group $\mathbb{R}^n$. This group-theoretic approach simplifies complex integral transformations.
  • Wavelet Multipliers: They define a specific operator class, $\phi T_\sigma \bar{\phi}$, where $\phi$ is an "admissible wavelet" and $\sigma$ is a "symbol"—a function that dictates how the operator acts in the frequency domain.
  • Riesz-Thorin Interpolation: To prove results for a wide range of spaces ($L^p$ for $1 < p < \infty$), the authors first establish properties for the extreme cases ($L^1$ and $L^\infty$) and then use interpolation theory to generalize the findings.
  • Fourier Analysis: The paper relies heavily on the Plancherel theorem and the fundamental relationship between convolution and multiplication in the Fourier domain.

Key Results and Insights

1. Boundedness and the Symbol-Class Correspondence

In Theorem 3.3, the authors prove that for a symbol $\sigma$ in $L^p(\mathbb{R}^n)$, there exists a unique bounded linear operator $P_\sigma$ on $L^2(\mathbb{R}^n)$. This establishes a direct link between the "size" of the symbol $\sigma$ (measured by its $L^p$ norm) and the behavior of the resulting operator.

2. The Schatten-von Neumann Property

Theorem 4.3 represents the core mathematical contribution of the paper. It states that if $\sigma \in L^p(\mathbb{R}^n)$ for $1 \leq p \leq \infty$, then the wavelet multiplier $\phi T_\sigma \bar{\phi}$ belongs to the Schatten-von Neumann class $S_p$. The authors provide a specific quantitative bound for the operator norm:

$$\|\phi T_\sigma \bar{\phi}\|_{S_p} \leq \frac{4^{1/p} \|\phi\|_\infty^{2/p'}}{(2\pi)^{n/p} \|\phi\|_4^4} \|\sigma\|_p$$

This result ensures that if the underlying symbol is well-behaved, the operator itself possesses the mathematical "niceness" required for stable computation and analysis.

3. The Landau-Pollak-Slepian Connection

The most significant insight for signal processing appears in Section 5. The authors demonstrate that the Landau-Pollak-Slepian operator—traditionally viewed as a sequence of projections onto time and frequency intervals—is actually a wavelet multiplier in disguise. Specifically, they prove that the wavelet multiplier is unitarily equivalent to the Landau-Pollak-Slepian operator $P_\Omega Q_T P_\Omega$ scaled by a factor of $1/2\Omega$. This means that the abstract theorems regarding $S_p$ classes apply directly to the tools used to analyze time- and band-limited signals.

Conclusion and Future Directions

He and Wong successfully unified abstract operator theory with practical signal analysis, showing that wavelet multipliers are a natural language for describing constrained signals. However, several avenues for future work remain:

  • Computational Implementation: As a theoretical work, the paper does not address the complexity of implementing these multipliers or their performance on noisy, real-world data.
  • Wavelet Selection: While the paper defines properties for "admissible wavelets," it primarily utilizes a specific characteristic function for the Landau-Pollak-Slepian connection. Future research could explore how different wavelet geometries (e.g., Gaussian or Meyer wavelets) influence the operator's properties.
  • Dimensionality: While the theory is formulated for $\mathbb{R}^n$, the practical examples are largely 1D. Extending these applications to image processing (2D) or volumetric data (3D) remains a natural next step.

Summary updated

Abstract

Abstract The Schatten-von Neumann property of a pseudo-differential operator is established by showing that the pseudo-differential operator is a multiplier defined by means of an admissible wavelet associated to a unitary representation of the additive group R n on the C * -algebra of all bounded linear operators from L 2 (R n ) into L 2 (R n ). A bounded linear operator on L 2 (R) arising in the Landau, Pollak and Slepian model in signal analysis is shown to be a wavelet multiplier studied in this paper.
BibTeX
@article{zhao1999wavelet,
  author = {He Zhao and M. W. Wong},
  title = {{Wavelet multipliers and signals}},
  journal = {The Journal of the Australian Mathematical Society Series B Applied Mathematics},
  year = {1999},
  doi = {10.1017/s0334270000010523},
}
Figure