Page 1 of 6
European Journal of Applied Sciences – Vol. 10, No. 6
Publication Date: December 25, 2022
DOI:10.14738/aivp.106.13526. Kitagawa, K. (2022). Prediction Method for Large Grasping Force using Single-Channel Surface Electromyography: A Pilot Study.
European Journal of Applied Sciences, 10(6). 320-325.
Services for Science and Education – United Kingdom
Prediction Method for Large Grasping Force using Single-Channel
Surface Electromyography: A Pilot Study
Kodai Kitagawa
Mechanical and Medical Engineering Course
Department of Industrial Systems Engineering
National Institute of Technology, Hachinohe College, Japan
ABSTRACT
Frequent grasping with a large force in the workplace leads to work-related
musculoskeletal disorders (WMSDs). Thus, the frequency and force of grasping in
each worker should be monitored for preventing these WMSDs. The frequency of
grasping can be counted by observation and existing methods using vision-based
devices or wearable sensors. Grasping force measurement generally requires force
sensors between the hand and objects. Almost force sensors are limited to the
surface and shape of the object in measurement; thus, measurement method
without force sensors is necessary to monitor grasping force in various workplaces.
Previous studies developed measurement methods for grasping force using
electromyography (EMG). However, almost existing methods require multiple- channels EMG for grasping force measurement. In addition, many methods were not
applied to large grasping forces. Therefore, the objective of this study is to develop
the prediction method for large grasping force using single-channel EMG. The
proposed method predicts grasping force by machine learning and features
calculated from single channel surface EMG (sEMG) signal on flexor digitorum
superficialis (FDS) muscle. As a pilot study, the proposed method was tested by
sEMG data of different large grasping forces (5 to 60 kgf). The results showed that
the correlation between the actual and predicted grasping force was greater than
0.8. These results indicate that the proposed method could measure large grasping
force by only single-channel sEMG.
Keywords: Grasping force; Surface electromyography; Occupational health; Machine
Learning.
INTRODUCTION
Grasping in the workplace leads to work-related musculoskeletal disorders (WMSDs) [1–3].
Especially, grasping tasks with high frequency and large force are known as a risk factor for
carpal tunnel syndrome (CTS) in occupational health [1]. In addition, there were high
prevalence ratios for WMSDs in high-frequency grasping with more than 4.0 kgf [2].
Furthermore, Shanahan et al. reported that grasping effort is an effective factor to evaluate
WMSDs risk in the upper limbs [3]. From these findings, it is considered that the frequency and
force of grasping in each worker should be monitored for preventing these WMSDs.
The frequency of grasping can be counted by observation and existing methods using vision- based devices or wearable sensors such as accelerometers, strain sensors, or electromyography
[4–6]. On the other hand, grasping force measurement generally requires force sensors
Page 2 of 6
321
Kitagawa, K. (2022). Prediction Method for Large Grasping Force using Single-Channel Surface Electromyography: A Pilot Study. European Journal
of Applied Sciences, 10(6). 320-325.
URL: http://dx.doi.org/10.14738/aivp.106.13526
between the hand and objects [7,8]. However, common force sensors are limited for the surface
and shape of the object in measurement [9]. Thus, new measurement methods without force
sensors are necessary. Previous studies developed measurement methods for grasping force
using electromyography (EMG) [10–16].
Table 1 shows the related studies of grasping force prediction using EMG [10–16]. Table 1
shows that the applied grasping forces of 4 related studies were less than 100 N (approximately
10 kgf) [11,12,14,15]. In other studies, applied grasping forces were larger than 100 N [10,13].
Existing methods of these studies might be used for preventing WMSDs due to large grasping
force; however, these studies required multiple-channels EMG [10,13]. Single-channel EMG
might be better for useful, comfortable, and continuous application in WMSDs prevention. On
the other hand, Kamavuako et al.’s study provided grasping force prediction using single- channel EMG in 2009 [15]. However, as mentioned previously, large grasping forces larger than
100 N were not applied in this study [15]. In addition, Kamavuako et al.’s further study provided
prediction method using single-channel EMG for grasping force larger than 100 N [16].
However, it is difficult to use this method for monitoring application because this method
requires invasive intramuscular EMG [16].
From these backgrounds, the objective of this study is to develop the prediction method for
large grasping force using single-channel non-invasive surface EMG (sEMG).
Table 1. Related Studies of Grasping Force Prediction using EMG
Reference Electrode Number of
Channel
Applied Grasping
Force [N]
Wu et al. [10] Surface 3 to 6 Maximum: 312.7
Kim et al. [11] Surface 6 Maximum: 40
Martinez et al. [12] Surface 4 to 16 Maximum: 31.1
Yang et al. [13] Surface 8 Maximum: 467
Ma et al. [14] Surface 16 Maximum: 92
Kamavuako et al. (2009)
[15] Surface Single Maximum: 50
Kamavuako et al. (2013)
[16] Intramuscular Single Mean ± S.D.: 481 ± 69
Proposed Method
Figure 1 shows the proposed method for predicting large grasping force using single-channel
sEMG. The proposed method predicts grasping force by machine learning and features
calculated from single channel sEMG signal on flexor digitorum superficialis (FDS) muscle. The
relationships between the grasping and FDS muscle were reported by many previous studies
[17–20]. Features (mean, median, maximum, minimum, S.D., variance, kurtosis, and skewness)
for machine learning-based regression were selected from related studies using machine
learning [21,22]. As a pilot study, this study evaluated whether the proposed method could
Page 3 of 6
322
European Journal of Applied Sciences (EJAS) Vol. 10, Issue 6, December-2022
Services for Science and Education – United Kingdom
predict large grasping force by only single-channel sEMG and machine learning. Furthermore,
artificial neural network (ANN), support vector machine (SVM), k-Nearest Neighbor (kNN),
gaussian process, linear regression, and M5P were compared to find a suitable machine
learning algorithm for regression of the proposed method.
Figure 1. Grasping Force Prediction
EXPERIMENT
This study evaluated whether the proposed method could predict large grasping force by only
single-channel sEMG and machine learning The participant of the experiment was one young
male (28 years, 189 cm, 168 kg). The participant provided informed consent to this experiment.
The participant performed grasping the adjustable hand-gripper with different grasping forces.
The posture of the participant was defined in Figure 1. Grasping forces were set as 5, 10, 20, 30,
40, 50, and 60 kgf by the adjustable hand-gripper. The participant performed 10 trials of
grasping for each grasping force. Grasping of each trial was kept for 5 seconds. The order of
trial and grasping force was randomized. Note that there was 2 minutes rest time after each
trial. Muscle activity of the FDS muscle during grasping was measured by surface EMG (EMG- EYE2, Global Linx Technology Co.,Ltd., Japan). The sampling rate of sEMG was 500 Hz.
Measured sEMG signals were processed via full-wave rectification. Features (mean, median,
maximum, minimum, S.D., variance, kurtosis, and skewness) for machine learning were
calculated from processed sEMG signals. These signal processing were performed by MATLAB
R2020b (Mathworks, Inc., Natick, MA).
Machine learning algorithms for the proposed method (ANN, SVM, kNN, gaussian process,
linear regression, and M5P) were implemented and tested in WEKA [23]. Training and testing
of machine learning were performed via leave-one-out cross-validation (LOOCV). Correlation