Towards Whole Body Fatigue Assessment of Human Movement: A Fatigue-Tracking System Based on Combined sEMG and Accelerometer Signals

Haiwei Dong, Izaskun Ugalde, Nadia Figueroa, Abdulmotaleb El-Saddik

TL;DR

This paper introduces a novel wearable system that combines sEMG and accelerometer data to track localized and overall human body fatigue during movement. By modeling fatigue as a dynamic process and fusing data from multiple muscles, the system offers a more comprehensive assessment than previous methods.

What Problem Does This Paper Address?

Monitoring human fatigue is crucial for preventing musculoskeletal disorders and optimizing performance in various fields, from sports training to rehabilitation. While localized muscle fatigue has been studied, assessing overall body fatigue during dynamic movements remains a challenge. Existing methods often treat fatigue as a static state or focus on individual muscles, failing to capture the complex, dynamic nature of fatigue across the entire body during complex movements. This paper aims to bridge this gap by developing a system that can dynamically track and quantify overall human movement fatigue.

Methods

The core of the proposed system lies in its multi-stage approach:

  • Signal Pre-processing: Raw sEMG and accelerometer data are filtered (Butterworth filter) to remove noise and then resampled to a common rate.
  • Periodic Movement Detection: The system uses cross-covariance analysis on the acceleration signal to identify periodic movements. This is crucial because the sEMG signal is only segmented and analyzed during active muscle contraction phases within these periodic movements. Non-periodic movements are handled differently, with the sEMG signal being processed directly.
  • Localized Fatigue Modeling: The paper leverages the established finding that the mean frequency of sEMG signals decreases linearly with muscle working time during fatigue. They statistically validate this linear relationship and define a localized fatigue level based on the relative decrease in mean frequency.
  • Dynamic Fatigue Tracking: To account for fatigue as a dynamic process, a "forgetting factor" is introduced. This factor allows the system to dynamically update fatigue parameters, giving more weight to recent measurements while still considering past fatigue information. This is implemented using a recursive least squares algorithm.
  • Overall Fatigue Fusion: Localized fatigue levels from different muscles are fused to compute an overall fatigue level. The fusion coefficients are determined by the rate of change (gradient) of each muscle's fatigue level, meaning muscles showing faster fatigue contribute more to the overall assessment.

Datasets

The developed fatigue-tracking system was evaluated using data collected from two fatigue experiments involving 17 subjects (10 males, 7 females). The experiments were:

  1. Dip start position training: Holding self-body weight with both arms.
  2. Arm curl training: Periodically lifting a weight (10 kg for males, 6 kg for females) with one arm.

No pre-existing datasets were used for training or evaluation; the data was collected specifically for this study.

Key Insights and Ideas

  • Fatigue as a Dynamic Process: The paper emphasizes that fatigue isn't static. By incorporating a "forgetting factor," the system dynamically updates fatigue levels, making it more responsive to real-time changes.
  • Fusion of Localized Fatigue: Combining fatigue information from multiple muscles, weighted by their rate of change, provides a more holistic view of overall body fatigue than looking at individual muscles in isolation.
  • Accelerometer for Movement Context: Using accelerometer data to detect periodic movements is a clever way to segment sEMG signals, ensuring that fatigue analysis is performed only during relevant muscle activity, thus improving accuracy and reducing noise.
  • Linearity of Mean Frequency Decrease: The statistical validation of the linear relationship between working time and mean frequency decrease provides a robust foundation for their localized fatigue model.

Key Results

  • The system successfully tracked the mean frequency trend of biceps brachii, converging to a negative slope within 5 seconds in both experiments, indicating physical fatigue.
  • In Experiment 1, localized fatigue levels for biceps brachii, anterior deltoids, and triceps brachii decreased to -37%, -9%, and -37% respectively, with an overall fatigue level of -32%.
  • In Experiment 2, localized fatigue levels reached -30%, -6%, and -28% for the same muscles, with an overall fatigue level of -17%.
  • A survey of the subjects indicated that the system's comfort level was generally rated as "moderately comfortable" to "very comfortable."
  • The computed fatigue levels were consistent with the subjects' self-reported fatigue and observed muscle-shaking phenomena.

Limitations and Future Work

  • Parameter Setting: The authors acknowledge that the "forgetting factor" and fatigue level fusion coefficients might need to be adjusted for different muscle types and applications. Clarifying these parameter settings is a key area for future work.
  • Subject Diversity: While the study involved 17 subjects from diverse backgrounds, further validation with a larger and more varied population would strengthen the findings.
  • Specific Movement Types: The current system is validated for two specific movements. Extending its applicability to a wider range of human movements would be beneficial.

Summary updated

BibTeX
@inproceedings{dong2014towards,
  author = {Haiwei Dong and Izaskun Ugalde and Nadia Figueroa and Abdulmotaleb El-Saddik},
  title = {{Towards Whole Body Fatigue Assessment of Human Movement: A Fatigue-Tracking System Based on Combined sEMG and Accelerometer Signals}},
  booktitle = {Sensors},
  year = {2014},
  url = {https://openreview.net/forum?id=WiIj4gaPUN},
}
Figure