<p>Surface electromyography (sEMG) has been
used to monitor muscle activity and predict fatigue in the workplaces. However,
objectively measuring fatigue is challenging in complex work with unpredictable
work cycles, where sEMG may be influenced by the dynamically changing posture
demands. The sEMG is affected by various variables and substantial change in
mean power frequencies (MPF), and a decline over 8-9% is primarily considered musculoskeletal
fatigue. These MPF thresholds have been frequently used, and there were limited
efforts to test their appropriateness in determining musculoskeletal fatigue in
live workplaces (which predominantly consist of complex tasks). In addition,
the techniques that consider both muscular and postural measurements that incorporate
dynamic posture changes observed in complex work have not yet been explored.
The overall objective of this work is to leverage both postural and muscular
cues to identify musculoskeletal fatigue in complex tasks/jobs (i.e., tasks
involving different levels of exertions, durations, and postures). The work was
completed in two studies.</p>
The first study aimed to
(1) predict subjective fatigue using objective measurements in non-repetitive
tasks, (2) determine whether the musculoskeletal fatigue thresholds in
non-repetitive tasks differed from the previously reported threshold, and (3)
utilize the empirically calculated thresholds to test their appropriateness in
determining musculoskeletal fatigue in live surgical workplaces. The findings
showed that the multi-modal measurements indicate better sensitivity than
single-modality (sEMG) measurements in detecting decreases in MPF, a predictor
of fatigue. In addition, the results showed that the thresholds in dynamic
non-repetitive tasks, like surgery, are different than the previously reported
8% threshold. Additionally, implementing muscle-specific thresholds increased
the likelihood of more accurately reporting subjective fatigue. The second
study aimed to develop a multi-modal fatigue index to detect musculoskeletal
fatigue. A controlled laboratory study was performed to simulate the
non-repetitive physical demands at different postures. A series of experiments
were conducted to test the effectiveness of
various metrics/models to identify subjective fatigue in complex tasks. Next, the
composite fatigue index (CFI) function was developed using the time-synced
integration of both muscular signals (measured with sEMG sensors) and postural
signals (measured with Inertial Measurement Unit (IMU) sensors). The variables
from sEMG (amplitude, frequency, and the number of muscles showing signs of
fatigue) and IMU (the prevalence of static and demanding postures and the number
of shoulders in static/demanding posture) sensors were integrated to generate
the CFI function. The prevalence of static/demanding postures was developed
using the cumulative exposures to static/demanding postures based on the material
fatigue failure theory. The single value fatigue index was obtained using the
resultant CFI function, which incorporates both muscular and postural
variables, to quantify the muscular fatigue in dynamic non-repetitive tasks.
The findings suggested that the propagation of musculoskeletal fatigue can be
detected using the multi-modal composite fatigue index in complex tasks. The
resultant CFI function was then applied to surgery tasks to differentiate the
fatigued and non-fatigued groups. The findings showed that the multi-modal
fatigue assessment techniques could be utilized to incorporate the muscular and
postural measurements to identify fatigue in complex tasks beyond
single-modality assessment approaches.
| Identifer | oai:union.ndltd.org:purdue.edu/oai:figshare.com:article/15164964 |
| Date | 13 August 2021 |
| Creators | Hamed Asadi (10875660) |
| Source Sets | Purdue University |
| Detected Language | English |
| Type | Text, Thesis |
| Rights | CC BY 4.0 |
| Relation | https://figshare.com/articles/thesis/Multi-Modal_Sensing_Approach_for_Objective_Assessment_of_Musculoskeletal_Fatigue_in_Complex_Work/15164964 |
Page generated in 0.0024 seconds