Fractional Fourier transforms on L and applications

Wei Chen, Zunwei Fu, Loukas Grafakos, Yue Wu

TL;DR

This paper extends the theory of the Fractional Fourier Transform (FRFT) to the $L^p(\mathbb{R})$ space for $1 \le p < 2$. It tackles convergence issues in $L^1$ by introducing a chirp operator and approximation methods for FRFT inversion, and establishes $L^p$ multiplier results and a Littlewood-Paley theorem in the FRFT context. This work is crucial for a deeper understanding and broader application of the FRFT in signal processing and other fields.

What Problem Does This Paper Address?

The classical Fourier Transform (FT) is a cornerstone of signal processing and analysis, but it struggles with non-stationary signals. The Fractional Fourier Transform (FRFT) offers a generalization that can better handle such signals by providing a tunable "time-frequency" representation. While the FRFT has been studied on $L^2(\mathbb{R})$ and Schwartz spaces, its behavior on $L^p(\mathbb{R})$ spaces, particularly for $1 \le p < 2$, has remained less explored. This paper aims to fill this gap by developing the $L^p$ theory of the FRFT, focusing on the challenges of convergence and inversion in the $L^1(\mathbb{R})$ setting, and extending these results to $L^p(\mathbb{R})$ for $1 < p < 2$.

Methods

The authors introduce a suitable chirp operator to analyze FRFT properties on $L^1(\mathbb{R})$. To address convergence issues and the FRFT inversion problem in $L^1(\mathbb{R})$, they employ approximation by suitable means, specifically fractional Gauss and Abel means. The paper also establishes the regularity of fractional convolution and derives results on the pointwise convergence of FRFT means. Furthermore, it delves into $L^p$ multiplier results and a Littlewood-Paley theorem within the FRFT framework. The FRFT on $L^p(\mathbb{R})$ for $1 < p < 2$ is defined by decomposing functions into $L^1$ and $L^2$ components.

Datasets

Not discussed in the paper.

Key Insights and Ideas

  • FRFT on $L^1(\mathbb{R})$ is tricky: Basic manipulations that work for $L^2$ can lead to convergence problems in $L^1$, making inversion non-trivial.
  • Chirp operators are key: Introducing chirp operators helps in understanding the structure of FRFT and its behavior, especially in $L^1$.
  • Approximation is the solution for inversion: Fractional Gauss and Abel means provide a way to recover the original function from its FRFT, even when the FRFT itself is not integrable.
  • FRFT generalizes classical concepts: The paper shows how concepts like approximate identities, convolution, and multipliers can be naturally extended to the FRFT.
  • $L^p$ theory for $1 < p < 2$ is built on $L^1$ and $L^2$: The FRFT for this range is defined by decomposing functions into $L^1$ and $L^2$ parts, leveraging the established theories for those spaces.

Key Results

  • The paper provides a rigorous definition and analysis of the FRFT on $L^1(\mathbb{R})$, including its properties and inversion via approximation methods.
  • It establishes the Riemann-Lebesgue lemma for the FRFT, showing that FRFTs of $L^1$ functions tend to zero at infinity.
  • The FRFT is shown to be a bounded linear operator from $L^1(\mathbb{R})$ to $L^\infty(\mathbb{R})$ and is uniformly continuous.
  • A multiplication formula for the FRFT is derived, which is crucial for proving other results.
  • The paper extends the FRFT to $L^p(\mathbb{R})$ for $1 < p < 2$ using a decomposition into $L^1$ and $L^2$ components, and proves the Hausdorff-Young inequality for FRFT multipliers.
  • A fractional Littlewood-Paley theorem is presented, providing a way to decompose functions based on the FRFT.
  • The paper demonstrates the effectiveness of FRFT in analyzing chirp signals, showing how the FRFT can recover a non-integrable chirp signal using approximation methods.

Limitations and Future Work

  • FRFT inversion on $L^1(\mathbb{R})$ is not always straightforward: While approximation methods are provided, the direct integral inversion may not always make sense if the FRFT is not integrable.
  • The paper focuses on $1 \le p < 2$: Extending the theory to $p \ge 2$ might require different techniques.
  • Applications are hinted at but not fully explored: The paper mentions applications in signal processing and optics, but a deeper dive into specific applications could be a future direction.
  • The authors suggest that FRFT inversion on $L^p(\mathbb{R})$ for $1 < p < 2$ can be proven using similar arguments as for $L^1(\mathbb{R})$, but do not elaborate on the details.
  • Future work could explore the FRFT's behavior on other function spaces or its applications in more complex signal processing scenarios.

Summary updated

BibTeX
@inproceedings{chen2021fractional,
  author = {Wei Chen and Zunwei Fu and Loukas Grafakos and Yue Wu},
  title = {{Fractional Fourier transforms on L and applications}},
  booktitle = {Applied and Computational Harmonic Analysis},
  year = {2021},
  doi = {10.1016/j.acha.2021.04.004},
}

Loading related content...

Figure