Hilbert-Huang Transform
Also known as: HHT, EMD-Hilbert Spectral Analysis, Hilbert Spektral Analizi, Adaptive Time-Frequency Decomposition
The Hilbert-Huang Transform (HHT) is an adaptive, data-driven method for analyzing non-linear and non-stationary time series, introduced by Norden E. Huang and colleagues in 1998. It combines Empirical Mode Decomposition (EMD), which decomposes a signal into intrinsic mode functions (IMFs), with the Hilbert spectral analysis to produce instantaneous frequency and amplitude representations without assuming signal stationarity or linearity.
Read the full method
Sign in with a free account to read this section.
Method map
The neighbourhood of related methods — select a node to explore.
When to use it
HHT is appropriate when the signal is known or suspected to be non-stationary (its statistical properties vary over time) or non-linear, such as biomedical signals, seismic waveforms, ocean wave data, or financial time series. It requires no prior basis functions and adapts fully to the data, making it superior to Fourier or wavelet methods when intrinsic oscillatory modes vary in frequency over time. It is not suitable when the signal is strictly stationary and linear, where Fourier analysis suffices with lower computational cost. Limitations include mode mixing, sensitivity to noise, and lack of a strict mathematical theory comparable to Fourier analysis.
Strengths & limitations
- Fully adaptive and data-driven: no pre-defined basis functions are assumed, making it suitable for non-linear and non-stationary signals.
- Provides true instantaneous frequency and amplitude estimates rather than time-averaged spectral content.
- Decomposes complex multi-component signals into physically meaningful intrinsic mode functions.
- Works directly on the data without requiring signal stationarity, linearity, or prior knowledge of signal structure.
- Mode mixing: different physical oscillations may appear in the same IMF, or a single oscillation may be split across multiple IMFs, especially in the presence of noise or intermittent signals.
- Lack of a rigorous mathematical theory: convergence and uniqueness of the EMD sifting process are not guaranteed in general.
- End effects: cubic spline interpolation of envelopes can introduce boundary distortions that propagate inward.
- Computationally intensive for very long time series and sensitive to the choice of sifting stoppage criterion and interpolation method.
Frequently asked
How does HHT differ from wavelet analysis?
Unlike wavelets, which use pre-defined mother wavelets and fixed time-frequency resolution trade-offs governed by the uncertainty principle, HHT derives its basis functions (IMFs) entirely from the data. This gives HHT adaptive frequency resolution, allowing it to track instantaneous frequency changes that wavelets may smear or miss. However, wavelets have a well-established mathematical framework while HHT's theoretical foundations remain an active research area.
What is mode mixing and how can it be addressed?
Mode mixing occurs when an IMF contains oscillations of widely different scales or when a single physical oscillation is spread across multiple IMFs, typically caused by intermittency or noise. Ensemble EMD (EEMD) addresses this by averaging IMFs obtained from multiple runs of EMD on noise-perturbed versions of the signal, causing noise-induced mode mixing to cancel out statistically while preserving the true signal modes.
Does HHT require any assumptions about the signal distribution?
HHT makes no assumptions about signal stationarity, linearity, or distributional form. It is entirely empirical and data-adaptive. The only implicit assumptions are that the signal can be meaningfully decomposed into oscillatory modes defined by extrema, and that cubic spline interpolation adequately represents the local mean envelope. This generality is the method's primary strength but also contributes to the lack of formal statistical inference tools.
Sources
- Huang, N. E., et al. (1998). The empirical mode decomposition and the Hilbert spectrum for nonlinear and non-stationary time series analysis. Proceedings of the Royal Society A, 454(1971), 903–995. DOI: 10.1098/rspa.1998.0193 ↗
How to cite this page
ScholarGate. (2026, June 2). Hilbert-Huang Transform. ScholarGate. https://scholargate.app/en/signal-processing/hilbert-huang-transform
Which method?
Set this method beside its closest kin and read them side by side — the library lays the books on the table; the choice is yours.
- Empirical Mode DecompositionSignal Processing↔ compare
- Fourier TransformSignal Processing↔ compare