Wigner–Ville distribution

Wigner–Ville distribution
Geek3 · CC BY-SA 4.0 · source
Name	Wigner–Ville distribution
Caption	Time–frequency representation
Introduced	1932 (Wigner), 1949 (Ville)
Application	Signal analysis, quantum mechanics, radar

Wigner–Ville distribution The Wigner–Ville distribution is a quadratic time–frequency representation introduced in the 20th century that links concepts from Eugene Wigner, Jean Ville, and quantum phase-space methods. It provides high-resolution localization of energy in time and frequency while connecting to formalisms used by Paul Dirac, Werner Heisenberg, Max Born, and John von Neumann. The distribution plays a central role across signal processing communities associated with institutions such as Bell Labs, Massachusetts Institute of Technology, Stanford University, École Normale Supérieure, and INRIA.

Definition and mathematical formulation

The Wigner–Ville distribution for a complex-valued signal x(t) is defined via a bilinear transform using time-shift and frequency-shift operators familiar from Hermann Weyl's quantization and the Heisenberg group. Formally it is expressed as an integral of the instantaneous autocorrelation x(t + τ/2) x*(t − τ/2) multiplied by a complex exponential, linking to the Fourier transform and the Stone–von Neumann theorem. Alternate formulations use operator kernels studied by Eugene Wigner and phase-space methods related to Weyl transform and Moyal bracket. The mathematical framework invokes distributions and tempered distributions as in the work of Laurent Schwartz and harmonic analysis developed by Antoni Zygmund.

Properties and interpretation

The Wigner–Ville distribution satisfies marginals that recover time-domain and frequency-domain energy measures via connections to the Plancherel theorem and the Parseval identity used by Norbert Wiener and John R. Pierce. It is real-valued and has time–frequency covariance under the action of the Heisenberg group and metaplectic operators associated with André Weil and Lionel Schwartz. Cross-terms arise due to bilinearity, producing interference artifacts analogous to quantum interference analyzed by Richard Feynman and phase-space negativities discussed by Roy Glauber. The distribution obeys energy conservation analogous to formulations by Paul Lévy and transforms under time and frequency shifts like representations considered by Hermann Minkowski in geometry.

Computation and numerical methods

Numerical evaluation typically uses discrete approximations related to the short-time Fourier transform implementations developed at Bell Labs and fast algorithms leveraging the fast Fourier transform by James Cooley and John Tukey. Discrete Wigner distributions require sampling theory connected to Claude Shannon and aliasing considerations studied by Harry Nyquist. Practical computation employs windowing, smoothing kernels, and discrete bilinear forms with algorithmic contributions from researchers at University of Cambridge, Imperial College London, and ETH Zurich. Regularization techniques leverage methods from numerical analysis advanced by Alan Turing and John von Neumann.

Relations to other time–frequency representations

The Wigner–Ville distribution relates to the spectrogram generated by the short-time Fourier transform and to Cohen's class of distributions associated with Leon Cohen. It connects to the ambiguity function and radar signal methods pioneered by Dennis Gabor and Frits Zernike, and to wavelet transforms developed by Yves Meyer, Stéphane Mallat, and Ingrid Daubechies. Cross-term behavior contrasts with linear representations used in techniques at Bell Labs and MIT Lincoln Laboratory, while smoothing kernels lead to distributions like the Choi–Williams distribution and Born–Jordan distribution studied in works associated with Max Born and Pascal Chopard.

Applications

Applications span audio signal analysis in research at IRCAM and Steinway & Sons collaborations; radar and sonar processing at Raytheon, BAE Systems, and Naval Research Laboratory; biomedical signal analysis in projects at Harvard Medical School and Johns Hopkins University; and quantum optics experiments at Caltech and CERN where phase-space representations of states draw on the original work of Eugene Wigner. Time–frequency feature extraction using the Wigner–Ville distribution appears in pattern recognition efforts at Carnegie Mellon University and machine learning pipelines influenced by researchers at Google Research and DeepMind.

Limitations and reassignment methods

The principal limitation is the existence of interference cross-terms that complicate interpretation, a problem noted in literature stemming from Leon Cohen and critics at IEEE. Mitigation includes smoothing kernels within Cohen's class, ridge extraction, and reassignment methods inspired by phase-space reassignment studied by F. Auger and groups at CNRS and EPFL. Reassignment and synchrosqueezing techniques relate to work by Daubechies and I. Daubechies' collaborators to sharpen representations while balancing bias and variance trade-offs addressed by Jerome Friedman and Bradley Efron.

Category:Time–frequency analysis