Fractional Fourier Transform (FrFT)

The FrFT represents a signal in a rotated time-frequency coordinate system. The transform order a (or rotation angle α = aπ/2) determines the degree of rotation. At a=0, the output is the original time-domain signal. At a=1, it becomes the classical Fourier transform. Intermediate orders provide representations that are partially in time and partially in frequency, making it ideal for separating components that overlap in time or frequency alone but are distinct in the rotated plane.

The FrFT kernel is composed of linear frequency modulated (LFM) chirp functions. This makes it the natural transform for analyzing chirp-like signals, which are ubiquitous in radar, sonar, and nature (e.g., bat echolocation). A monocomponent chirp signal becomes an impulse (delta function) in the appropriate fractional Fourier domain, enabling perfect compact representation and filtering that is impossible in either the pure time or pure frequency domains.

The FrFT obeys the index additivity property: applying an FrFT of order a₁ followed by an FrFT of order a₂ is exactly equivalent to a single FrFT of order a₁ + a₂. This forms a continuous, one-parameter unitary group of transformations. This property is crucial for iterative algorithms and ensures that the transform is a true rotation operator in the time-frequency plane, maintaining energy conservation (Parseval's theorem) for all orders.

The FrFT has a fundamental connection to the Wigner-Ville Distribution (WVD). Computing the squared magnitude of the FrFT of a signal is equivalent to projecting the signal's WVD onto the rotated axis and integrating. This means the FrFT directly performs Radon transform-like tomography on the time-frequency plane. This relationship is exploited for optimal filtering: by rotating to a domain where signal and noise do not overlap, applying a simple mask, and rotating back.

A discrete FrFT (DFrFT) can be computed by finding the eigenvectors of the discrete Fourier transform (DFT) matrix and raising the corresponding eigenvalues to the fractional power a. This approach ensures the discrete transform closely approximates the continuous FrFT's properties, including index additivity and approximation of the Hermite-Gaussian functions as eigenvectors. Fast O(N log N) algorithms exist, making it practical for real-time signal processing applications.

For signals with multiple chirp components having different sweep rates, the FrFT enables sequential component excision. By scanning through fractional orders a and computing the FrFT, each component will produce a distinct energy peak at a unique order and coordinate. This allows for:

• Blind source separation of overlapping chirps
• Interference excision in direct-sequence spread spectrum systems
• Feature extraction for automatic modulation classification in cognitive radio

The foundational method for time-frequency analysis that computes the Fourier transform on windowed segments of a signal. Unlike the FrFT's rotated time-frequency plane, the STFT uses a fixed rectangular tiling of the plane.

• Uses a fixed window function (e.g., Hamming, Gaussian)
• Resolution trade-off governed by the Heisenberg-Gabor uncertainty principle
• Spectrogram is the squared magnitude of the STFT
• Serves as the baseline against which FrFT's chirp-based decomposition is compared

A quadratic time-frequency distribution that provides the maximum possible joint resolution by computing the Fourier transform of the signal's instantaneous autocorrelation. The FrFT shares a deep mathematical connection with the WVD through the Radon transform relationship.

• The FrFT of a signal corresponds to rotating its WVD by the transform angle
• Suffers from cross-term interference for multi-component signals
• Provides the theoretical foundation for understanding FrFT as a rotation operator

A member of Cohen's class of distributions that applies an exponential kernel in the ambiguity domain to suppress cross-term interference while preserving auto-term resolution. Like the FrFT, it addresses the limitations of the WVD for multi-component signals.

• Exponential kernel: Φ(θ,τ) = exp(-θ²τ²/σ)
• Balances cross-term suppression with time-frequency concentration
• Used extensively in RF fingerprinting for clean spectral representations
• Complements FrFT-based chirp rate estimation in emitter identification pipelines

The time derivative of the instantaneous phase of an analytic signal, representing the dominant frequency at each moment. The FrFT excels at analyzing signals with linearly varying instantaneous frequency (chirps).

• Computed from the analytic signal via the Hilbert transform
• For a linear chirp, instantaneous frequency is a linear function of time
• The optimal FrFT order compacts the chirp into a delta function
• Critical for extracting modulation fingerprints from frequency-agile emitters

A reassignment technique that sharpens time-frequency representations by reallocating coefficients along the frequency axis based on instantaneous frequency estimates. Both SST and FrFT aim to improve component concentration.

• Operates on CWT or STFT outputs
• Concentrates energy along time-frequency ridges
• Enables mode extraction and reconstruction
• Used alongside FrFT for separating overlapping chirp components in radar and communications signals

A three-parameter unitary integral transform that generalizes the FrFT, Fresnel transform, and scaling operations. The FrFT is a one-parameter special case of the LCT.

• Parameterized by a 2×2 matrix with unit determinant
• Represents general linear symplectic transformations of the time-frequency plane
• Encompasses chirp multiplication, convolution, scaling, and Fourier transformation
• Provides the most general framework for optical and signal processing systems