Explorations in Time-Frequency Analysis
An authoritative exposition of the methods at the heart of modern non-stationary signal processing from a recognised leader in the field. Offering a global view that favours interpretations and historical perspectives, it explores the basic concepts of time-frequency analysis, and examines the most recent results and developments in the field in the context of existing, lesser-known approaches. Several example waveform families from bioacoustics, mathematics and physics are examined in detail, with the methods for their analysis explained using a wealth of illustrative examples. Methods are discussed in terms of analysis, geometry and statistics. This is an excellent resource for anyone wanting to understand the 'why and how' of important methodological developments in time-frequency analysis, including academics and graduate students in signal processing and applied mathematics, as well as application-oriented scientists.
Why Read This Book
You will gain a deep, unified view of modern time–frequency methods that explains not only how algorithms work but why they were developed and when to use them. Flandrin blends historical context, geometric intuition, and statistical insight so you can interpret results reliably across audio, speech, radar and communications applications.
Who Will Benefit
Advanced graduate students, researchers, and signal-processing engineers working on non-stationary signals (audio/speech, radar, communications, bioacoustics) who need a principled understanding of time–frequency methods and their limitations.
Level: Advanced — Prerequisites: Undergraduate-level signals and systems, Fourier analysis, linear algebra, and basic probability/statistics; familiarity with DSP concepts such as filters, STFT/FFT and basic wavelet ideas is highly recommended.
Key Takeaways
- Understand the theoretical foundations and interpretations of major time–frequency representations (STFT, Wigner–Ville, Cohen's class, wavelets).
- Implement and evaluate practical algorithms for time–frequency analysis, including reassignment and synchrosqueezing, with awareness of numerical trade-offs.
- Compare and choose appropriate methods for analyzing non‑stationary signals based on resolution, cross‑term behavior, and statistical robustness.
- Apply geometric and statistical perspectives to interpret time–frequency distributions and to assess significance in noisy or multi‑component signals.
- Analyze real-world waveform families (bioacoustics, speech, physics examples) using the covered methods and diagnose common failure modes.
Topics Covered
- Introduction and historical perspectives on time–frequency analysis
- Fundamental concepts: time–frequency localization and uncertainty
- Linear representations: STFT, spectrograms and continuous wavelet transform
- Quadratic/bilinear representations and Cohen's class
- The Wigner–Ville distribution: properties, cross‑terms and mitigation
- Reassignment methods and time–frequency sharpening
- Synchrosqueezing, mode extraction and source separation
- Wavelets and multiresolution approaches in time–frequency
- Discrete implementations, FFT usage and numerical considerations
- Statistical approaches: estimation, significance testing and noise effects
- Geometry of time–frequency representations and feature interpretation
- Applications: audio/speech, bioacoustics, radar and communications
- Recent developments, practical guidelines and open problems
- Appendices: mathematical tools and proofs
Languages, Platforms & Tools
How It Compares
Covers similar ground to Leon Cohen's classic on time–frequency distributions and Stéphane Mallat's Wavelet Tour, but Flandrin emphasizes historical context, geometric/statistical interpretation and recent techniques (e.g., synchrosqueezing) rather than exhaustive toolbox recipes.
