The Study on the sEMG Signal Characteristics of Muscular Fatigue Based on the Hilbert-Huang Transform

Bo Peng, Xiaogang Jin, Yong Min, Xianchuang SuLecture Notes in Computer Science 2006 · DOI · OpenAlex

Abstract

Monitoring muscular fatigue is essential for sports science, rehabilitation, and ergonomics. Traditional analysis of surface electromyography (sEMG) signals often relies on Fourier-based methods, which assume signal stationarity—a condition rarely met during muscle strain. This paper proposes the Hilbert-Huang Transform (HHT) as a more precise, adaptive tool for analyzing non-linear and non-stationary sEMG data. By applying Empirical Mode Decomposition (EMD) and the Hilbert Transform, the researchers identified a distinct "fatigue signature": a measurable drop in instantaneous frequency accompanied by a significant rise in instantaneous energy. These findings suggest that HHT provides a high-resolution time-frequency distribution that can pinpoint the onset of muscle failure more effectively than traditional averages.

Introduction

Monitoring muscular fatigue is a cornerstone of sports science and clinical rehabilitation. Surface electromyography (sEMG) is the standard non-invasive method for recording electrical muscle activity. However, sEMG signals are notoriously difficult to analyze because they are inherently non-linear and non-stationary.

Standard tools like the Fast Fourier Transform (FFT) assume that a signal's statistical properties remain constant over time. While indicators like Median Frequency (MF) are useful for static loads, they often fail to capture the dynamic shifts occurring as a muscle nears exhaustion. This study addresses these limitations by introducing the Hilbert-Huang Transform (HHT), a method specifically designed to extract meaningful features from messy, real-world biological data.

Methodology

The core innovation of this research is the application of HHT to the sEMG domain. The process involves two primary stages:

  1. Empirical Mode Decomposition (EMD): This adaptive "sifting" process breaks the signal down into Intrinsic Mode Functions (IMFs). Unlike Fourier or Wavelet transforms, EMD does not rely on pre-defined basis functions; instead, it decomposes the signal based on its own local characteristic time scales.
  2. Hilbert Transform: Once decomposed, the Hilbert Transform is applied to each IMF to calculate instantaneous frequency (IF) and instantaneous energy (IE) at every point in time.

Experimental Setup

The study involved 12 healthy male university students (ages 19–24). Subjects performed a static biceps contraction at 80% of their Maximal Voluntary Contraction (MVC) until total exhaustion. Signals were recorded at a sampling frequency of 1000Hz to ensure high-fidelity data for the HHT algorithm.

Results

The HHT analysis revealed a clear qualitative and quantitative shift in muscle behavior as exhaustion set in. The researchers compared the initial "fresh" state of the muscle to the final "exhausted" state, identifying two primary trends:

  • Frequency Drop: The instantaneous frequency showed a significant downward trend throughout the fatiguing process. In the final moments of the task, the frequency was concentrated in a much lower band compared to the start.
  • Energy Surge: Conversely, the instantaneous energy approximately doubled or tripled as the muscle fatigued, reflecting the increased neural drive required to maintain the contraction.
Trends in Instantaneous Energy (a) and Instantaneous Frequency (b) during fatiguing contraction.

Visually, the Hilbert Spectrum provides a high-resolution energy-frequency-time distribution. As shown in the comparison between the initial and final stages of contraction, the signal becomes "brighter" (indicating higher energy) and shifts toward the lower end of the frequency spectrum as time progresses.

Hilbert Spectra comparison between the initial (a) and final (b) stages of muscle contraction.

Discussion and Conclusion

This study demonstrates that HHT is a powerful, self-adaptive method for sEMG analysis. By letting the data "speak for itself," HHT avoids the resolution trade-offs inherent in Fourier or Wavelet methods. The resulting Hilbert Spectrum offers a more evident and convenient indicator of fatigue than traditional mean or median frequency averages.

However, the authors note that this is a preliminary exploration. The HHT method has known mathematical deficiencies, such as "mode mixing," which require further algorithmic refinement. Additionally, while the results are promising for static contractions of the biceps, future work must extend this analysis to dynamic movements and larger, more diverse populations to validate its utility for clinical or real-time wearable applications.

Summary updated

BibTeX
@inproceedings{peng2006study,
  author = {Bo Peng and Xiaogang Jin and Yong Min and Xianchuang Su},
  title = {{The Study on the sEMG Signal Characteristics of Muscular Fatigue Based on the Hilbert-Huang Transform}},
  booktitle = {Lecture Notes in Computer Science},
  year = {2006},
  doi = {10.1007/11758501\_23},
}

Loading related content...

Figure