Characterization of sEMG Spectral Properties During Lower Limb

Muscle Activation

Costa-Garcia Alvaro

and Shimoda Shingo

National Institute of Advanced Industrial Science and Technology (AIST) Kashiwa II Campus, University of Tokyo,

6-2-3 Kashiwanoha, Kashiwa, Chiba 277-0882, Japan

Nagoya University Graduate School of Medicine 64 Tsutumai, Showa-ku, Nagoya 466-8550, Japan

Keywords: Electromyography, Muscle Contraction Types, Artifacts.

Abstract: The analysis of biological data is an effective way to extract implicit information about the human

physiological condition, representing the performance of daily tasks. The use of this information as feedback

for robotic systems can contribute to a smoother transition into societies with a higher level of human-robot

collaboration. Superficial electromyography (sEMG) could be a powerful ally in this field, as muscle activity

serves as a window into our neural system and can be measured non-invasively with relative ease. In this

work, our objective is to extract spectral features that enable the classification between isometric and isotonic

muscle contractions. The switching between these types of contractions during human motion has been widely

linked to various physical conditions, such as muscle pain, fall prediction, postural imbalances, and stress. To

achieve this goal, we recorded muscle activity during both isometric and isotonic contractions under various

conditions. We conducted a time-frequency analysis on the data collected from five lower limb muscles of

four healthy subjects to extract significantly relevant features containing the necessary information to

discriminate between these two types of muscle activations. Our results suggest that this discrimination can

be achieved through the analysis of two spectral features: the median frequency and the power contained in

the frequency range between 11 and 32 Hz. Furthermore, the inclusion of the peak frequency as a third feature

also enables the detection of low-frequency motion artifacts.

1 INTRODUCTION

The development of increasingly efficient and

intuitive systems has enabled the progressive

adoption of new technologies by large population

groups. This phenomenon has opened the doors to an

era in which the integration between human and

technological systems will occur at a fast pace

(Ritchie, 2017; Besley, 1993). Since the COVID-19

pandemic and the burden it placed on medical centers

around the world, governments and health institutions

have been promoting the so-called 'digital health era,'

encouraging people to embrace new digital media

technologies for health self-monitoring. In this

direction, the integration of robotic solutions that

facilitate the self-monitoring of health conditions for

a wider population has been widely proposed by the

scientific community (Ahmed, 2021; Yang, 2019).

Furthermore, over the last few decades, there has

https://orcid.org/0000-0003-0097-2793

https://orcid.org/0000-0002-7759-7541

been an explosive increase in the development of

robotic systems for human support and augmentation

(Green, 2008; Shimoda, 2022). Given the

fundamental differences between the evolution-based

appearance of humans and the engineering-based

development of robotic systems, one of the biggest

challenges in this field is finding appropriate methods

for a smooth integration between these two natures.

To achieve harmonious collaboration, it is necessary

to establish communication channels between human

and robotic systems that allow for a certain level of

mutual understanding. Current scientific efforts in

this direction involve the use of biological signals as

control commands, feedback, and indicators for

robotic systems (Bainbridge, 2021). A notable

example of this approach is the Smart Wearable

Robot with Bioinspired Sensory-Motor Skill

(BioMot) project (Bacek, 2017; Costa, 2016), a

European project developed between 2013 and 2016.

Alvaro, C. and Shingo, S.

Characterization of sEMG Spectral Properties During Lower Limb Muscle Activation.

DOI: 10.5220/0012305000003657

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 17th International Joint Conference on Biomedical Engineering Systems and Technologies (BIOSTEC 2024) - Volume 1, pages 705-712

ISBN: 978-989-758-688-0; ISSN: 2184-4305

705

This project aimed to develop a lower limb

exoskeleton that uses kinetic data, muscle activity,

and real-time brain signals recorded from spinal cord

injury patients to adapt their rehabilitation therapies

to their current physical and cognitive state.

The use of muscle activity recorded from

superficial electromyography (sEMG) has shown

high performance when compared to other human

recorded data (Li, 2020). Among the different bio-

electrical signals, sEMG data contains more

information about human behavior (both motor and

neural data) than electrocardiogram (ECG) or

electrooculography (EoG), and its measurement can

be done with increasingly affordable and easy-to-set-

up systems compared to brain signal recordings like

electroencephalography (EEG). sEMG signals also

exhibit a higher signal-to-noise ratio when compared

to the latter systems. However, the challenge of

artifact coupling on sEMG recordings during motion

still needs to be addressed before their effective

integration into robotic systems aimed at daily human

support (Lienhard, 2015). In this regard, the current

paper focuses on characterizing sEMG spectral

properties under different lower limb motions to

establish a ground truth that can be used in future

research when these signals are employed as

biofeedback communication channels for robotic

devices (Lünenburger, 2007).

Therefore, the spectral characterization presented

in this paper has a dual objective. On the technical

side, the authors aim to highlight spectral features that

will enable future researchers to detect and remove

sEMG data coupled with motion artifacts, preventing

this contamination from affecting the human-robot

interaction stage. On the functional side, the goal is to

provide a set of features that allow a classification

system to distinguish between fundamentally

different human motions, enabling this information to

be easily shared in real-time with external devices.

For this purpose, the current research measures

sEMG signals during both isometric and isotonic

motions under different environmental conditions

where the coupling of motion artifacts is common.

Data are recorded from five different leg muscles

during regular walking tasks and are separated into

different pairs of motor and noise conditions. A time-

frequency analysis based on the Fast-Fourier

Transform was used to extract the spectral

distribution associated with each task. Comparing

spectral distributions between paired tasks allowed

for the extraction of significant differences between

experimental conditions. Finally, the observed

changes in spectral distribution were used to select

those features that would be most helpful in

differentiating between muscle contraction types and

identifying noisy data during daily motions.

In the following section, the authors introduce the

materials and methods used during this research. This

information includes details about the volunteers

participating in the experiment and technical

information about the experimental protocol, data

recording, data processing, feature extraction, and

analysis methodology.

2 MATERIALS AND METHODS

2.1 Participants

Four participants, consisting of 2 men and 2 women,

participated in the experiment. Their ages ranged

from 27 to 46 years old, with an average age of 36.25

± 8.42. All participants were right-footed and had no

history of motor diseases. They were fully informed

about the experimental conditions and provided their

informed consent in accordance with the Declaration

of Helsinki. The study was also approved by the

ethical review board of the RIKEN research institute,

with the ethical approval code: Wako3 28–13.

Figure 1: Experimental Conditions and Queue. A)

Ground walking. B) Treadmill walking. C) Isometric

contraction of gastrocnemius and vaslus lateralis muscle in

the top image, tibialis anterior in the middle image and

peroneus longus in the bottom image. D) Combination of

band and no band fixation condition with the usage or not

of footwear. E) Graphical representation of an experimental

session composed of 3 repetitions of 12 tasks resulting form

the combonantion of motion types, band fixations and

foowear usage.

BIOSIGNALS 2024 - 17th International Conference on Bio-inspired Systems and Signal Processing

706

Figure 2: Spectral Representation. Boxplot representation

of the spectral features extracted from a given experimental

condition. The boxplot distribution of paired task were

compared frequency by frequency using a Wilcoxon Sum-

rank Test to find statistical differences between conditions.

2.2 Experimental Protocol

Each participant completed a single experimental

session consisting of three repetitions of the task

sequence depicted in Figure 1. Each task was defined

as a combination of a motion condition (isometric

contraction, ground walking, and treadmill walking -

as shown in Figure 1A-C), electrode fixation system

(using a band for electrode fixation or not - as

depicted in Figure 1D), and footwear usage (with or

without footwear - as illustrated in Figure 1E). This

resulted in a total of 12 tasks per repetition, as

indicated in Figure 1E. During isotonic motions

(treadmill and ground walking conditions),

participants were asked to walk for 10 steps, while

during isometric contraction tasks, they were

instructed to activate their muscles for 10 seconds.

After completing the three repetitions, the sEMG data

included muscle activation data from 30 steps for

each task performed under isotonic motion and 30

seconds of data from tasks recorded during isometric

contractions. The inclusion of different electrode

fixation conditions aimed to assess motion artifacts

related to the vibration of hanging electrodes caused

by the momentum generated in the lower limb during

gait. Additionally, the use of footwear, along with the

choice between a treadmill or regular ground

walking, was considered to account for spectral

changes associated with the coupling of power line

noise originating from the environment or other

external devices.

2.3 sEMG Recordings

Muscle activity was recorded using five wireless

bipolar electrodes (BTS FREEEMG; BTS

Bioengineering Corp., Milan, Italy) placed on the

following muscles: peroneus longus (PL), tibialis

anterior (TA), vastus lateralis (VLAT),

gastrocnemius medialis (GAN), and gastrocnemius

lateralis (GAL). The placement of electrodes

followed the guidelines established by the Surface

Electromyography for the Non-Invasive Assessment

of Muscles (SENIAM) project (Stegeman, 2007).

Data were digitized at a rate of 1000 Hz.

The selection of these muscles was based on

criteria targeting muscles that are highly active during

walking motions (Costa, 2021). These muscles were

chosen because they are located in the lower part of

the legs, where motion momentum is higher, and

therefore, motion artifacts are also expected to be

stronger.

To ensure the correct activation of the five

selected muscles during all tasks, isometric

contraction conditions were recorded under two

different exercises, as shown in Figure 1C. The top

image in Figure 1C illustrates the exercise used for

the isometric activation of the GAN, GAL, and

VLAT muscles, while the lower images in Figure 1C

depict the position and exercise used for the TA and

PL isometric contraction.

2.4 Data Processing and Feature

Extraction

In this study, one of the primary objectives was to

evaluate the effects of artifacts on the spectral

properties of sEMG data. To ensure that the impact of

motion artifacts was included in the analysis, the

recorded signals did not undergo rectification or

filtering.

Before performing spectral computations, each

task was segmented into 10 epochs. The segmentation

techniques applied to isotonic and isometric tasks

differed due to their fundamental nature. For

isometric contractions (where 10 seconds of data

were extracted for each task), epochs were obtained

as consecutive one-second segments. In the case of

isotonic motion tasks, each epoch corresponded to the

muscle activation produced by a single step. The

starting and ending values of each segment were

determined independently for each muscle using a

methodology for periodic motion segmentation

previously introduced in (Costa, 2020). This method

identifies minima and maxima in the sEMG envelope

that best align with the expected number of

activations (10 steps per task in this study). It uses an

iterative low-pass filtering process to adjust the cutoff

frequency until the envelope synchronizes with the

number of steps. The extracted points serve as

segmentation markers for epoching the sEMG raw

segment data.

Characterization of sEMG Spectral Properties During Lower Limb Muscle Activation

707

Figure 3: Spectral differences across conditions. Figures from A to E show the frequency range in which statistical

significance was found among different pair of condition. Vertically arranged graph show the results for each one of the five

muscle recorded during the experiment. Red and green lines were used to show subdivision among the conditions compared

and blue lines represent average values.

BIOSIGNALS 2024 - 17th International Conference on Bio-inspired Systems and Signal Processing

708

Next, a Fast Fourier Transformation (FFT) was

applied to each raw epoch to extract their spectral

features. Additionally, the power at each frequency

was divided by the total power within the range of 1-

500 Hz, respecting the Nyquist criteria for not

aliasing frequencies (Robinson, 1991). This

normalization scales the total power of the spectrum

to 1, highlighting how power is distributed within the

spectrum by removing amplitude variations. This

normalization allows for better comparisons between

tasks and subjects in terms of spectral distribution.

Due to the observed spectral differences between

the various conditions of the analyzed data (for more

details, refer to the Results section), three distinct

features were extracted from each normalized

spectrum: a) peak frequency: this represents the

single frequency value containing the highest

amplitude in the spectrum; b) median frequency: This

indicates the frequency value that divides the

spectrum into two areas with equal power; and c)

power within the range of 11 to 32 Hz: This measures

the power contained within this specific frequency

range.

Figure 4: Spectral features. Bidimensional representation

of the 3 spectral features extracted from the main spectral

distributions. X-axis represents median frequency and Y-

axis represents the normalized amplitude in the range

between 11-32 Hz and the frequency recording the peak

amplitude respectively. Feature were painted with three

different colors to differentiate between isometric motions

(black dots), isometric motion without (red dots) and with

(blue) electrode fixation. A) Represent the feature extracted

for each subject independently. B) Shows the average data

for all participants.

Figure 5: Summary of main spectral distribution.

Average values associated to the main spectral distribution

found after sEMG analysis. A) Isometric contraction (left

graph), isotonic contraction during electrode fixation

(middle graph) and isotonic contraction without electrode

fixation (right graph). B) Comparison between main

spectral distributions

2.5 Data Comparison

The various tasks recorded during a single session

were organized into different groups to facilitate

comparisons between different experimental

conditions. These groupings encompassed

comparisons between isometric and isotonic tasks,

ground walking and treadmill walking tasks, tasks

with and without a band for electrode fixation, and

tasks with and without the use of footwear.

To identify statistically significant differences in

the spectrum between pairs of conditions, all the

spectra associated with the same condition were

represented using boxplots, as illustrated in Figure

2A. The values used for the boxplot representation at

each frequency were then compared between paired

tasks using a Wilcoxon sum-rank test, followed by a

Bonferroni-Holm correction to account for multiple

comparisons (Rey, 2011; Abdi, 2010). Subsequently,

the frequency values that exhibited statistical

significance were represented as the difference

between the median values of the respective tasks.

3 RESULTS

3.1 Spectral Shape Comparison

Figure 3 presents the results of spectral comparisons

for various paired conditions. In Figure 3A, it

becomes evident that the use of band fixation

significantly reduces low-frequency activation,

typically associated with motion artifacts. This

Characterization of sEMG Spectral Properties During Lower Limb Muscle Activation

709

reduction is particularly pronounced during isotonic

tasks but has minimal impact during isometric

contractions. In Figure 3B, the comparison between

tasks with and without footwear shows less

pronounced spectral differences, primarily affecting

the no-band fixation conditions. This suggests that

proper electrode fixation is more critical than the type

of footwear used.

Figures 3C-D compare isotonic and isometric

tasks. Figure 3C includes both band and no-band

fixation conditions, revealing an increased effect on

lower-frequency activation associated with the no-

band condition. When the no-band condition is

removed (Figure 3D), it becomes apparent that

isotonic motions lead to an expansion in the spectral

range between 20-140 Hz compared to isometric

motions. In this case, there are no notable differences

between ground walking and treadmill walking.

Furthermore, Figure 3E illustrates the spectral

differences between ground and treadmill walking

conditions. It is noticeable that an increase in low

frequencies during ground walking occurs only under

the no-band fixation condition, further supporting the

idea that when electrodes are securely fixed, there are

no significant spectral differences between ground

and treadmill walking.

Lastly, Figure 4A provides a summary of the main

spectral differences observed among the analyzed

conditions. The leftmost graph depicts the spectral

distribution of isometric tasks for the five evaluated

muscles, revealing no significant changes related to

footwear or fixation conditions. The middle graph

displays the spectral distribution of isotonic tasks

during electrode fixation conditions, showing no

significant differences between ground/treadmill

walking or footwear usage. The right graph

demonstrates the effects of not using a band for

electrode fixation in the low frequencies of the

spectrum. Electrodes 4 and 5, located in the

gastrocnemius muscle with less momentum, appear

less affected by lower frequency increases, further

supporting the hypothesis that this activation is a

consequence of motion artifacts. Figure 4B presents a

comparison between the three previous spectra (left

graph) and between isotonic and isometric conditions

when motion artifact conditions are not included

(right graph).

3.2 Feature Representation for

Classification

In Figure 5B, two bidimensional graphs provide a

comparison of the three extracted spectral features

(peak frequency, median frequency, and normalized

amplitude within the 11-32 Hz range) across different

conditions. The right graph illustrates the relationship

between peak and median frequencies, while the left

graph compares median frequency with the power

distribution in the 11-32 Hz range. Red dots represent

features extracted from isotonic motions where

electrodes were not fixed by the band. Blue dots show

the features of isotonic motions with properly fixed

electrodes. Finally, black dots represent features

extracted during isometric motions.

Figure 5A presents the same features, but they are

separated for each participant, which helps

underscore the level of inter-subject variability in the

analyzed data.

4 DISCUSSION

Our results reveal two primary changes in the spectral

distribution of sEMG data under the evaluated

conditions. The first change involves an increase in

the power of low frequencies observed during

isotonic data recordings without electrode fixation.

This change is clearly depicted in Figure 3A (red

dotted line) and Figure 3C (green dotted line).

Importantly, this phenomenon is absent during

isometric contractions (Figure 3C, red dotted line),

suggesting that the rise in lower frequencies is a

consequence of motion artifacts stemming from

electrode vibration during walking tasks. This

frequency alteration affects the range between 11 and

32 Hz, with a peak value at approximately 22 Hz,

aligning with the frequency range traditionally

associated with motion artifacts (Lienhard, 2015;

Fratini, 2009).

The second change involves a more gradual shift

between high and low frequencies when comparing

isometric and isotonic contractions, particularly

under electrode fixation conditions (Figure 3D). This

phenomenon indicates that, for isotonic motions,

there is an increase in the range between 30 and 100

Hz, compensated by a decrease between 200-300 Hz.

Unlike the previous range strongly linked to motion

artifacts, this affected range is much broader and has

minimal overlap. Furthermore, the frequencies

impacted fall within the range at which motor unit

action potentials are generated (Costa, 2022). This

suggests that the spectral differences between

isotonic and isometric contractions arise from

physiological differences rather than noise coupling.

Under band fixation conditions, no significant

changes were observed between footwear conditions

(Figure 3B, red dotted line) or between ground and

treadmill walking (Figure 3D, red dotted line). This

BIOSIGNALS 2024 - 17th International Conference on Bio-inspired Systems and Signal Processing

710

indicates that the coupling of external noise sources,

such as power line interference, has minimal

influence on the spectral distribution of sEMG,

making it easier to mitigate during recordings in daily

environments.

These findings enable the classification of sEMG

spectral distributions into three main groups:

isometric contractions, normal isotonic contractions,

and isotonic contractions with coupled motion

artifacts. The spectral features of each group exhibit

sufficient distinctiveness, as illustrated in Figure 5

(median frequency, peak frequency, and normalized

amplitude in the 11-32 Hz range). The black, blue,

and red feature clusters presented in Figure 5 suggest

that differentiation should be feasible with

straightforward classification techniques, enabling

real-time discrimination between isometric/isotonic

contractions and noisy/not noisy trials. Furthermore,

our results indicate low inter-subject variability in the

feature space (Figure 5A), which enhances the

potential for wider generalization of the classification

algorithm.

5 CONCLUSIONS

This study has identified distinct spectral features in

the sEMG spectrum that enable two important

outcomes: a) the discrimination between isotonic and

isometric contractions, and b) the detection of low-

frequency motion artifacts during walking tasks.

The differentiation between isometric and isotonic

contractions has widespread applications in motor

control, offering insights into various cognitive states

such as pre-fall postural instability (Xi, 2017),

motion-related muscle and joint pain (Neblett, 2016),

stress/anxiety-related head pain and migraines

(Bakal, 1977), among others. Consequently,

distinguishing between these fundamental

contractions represents a crucial initial step in

providing robotic systems with valuable information

about the cognitive state of the human. Additionally,

the ability to identify signal segments contaminated

by motion artifacts allows robotic devices to

determine the reliability of received physiological

data. As mentioned in the introduction, the challenge

of motion artifact coupling is pertinent to the

measurement of biosignals during human-robot

interaction. Precisely detecting noisy motion trials

will aid in assessing current integrative solutions and

developing innovative approaches to noise reduction.

Furthermore, comparing the spectral distributions

obtained in this study can facilitate future research in

evaluating changes in sEMG recordings following the

integration of robotic devices into the human control

loop. After further validation through a larger subject

sample, future steps will involve developing a

classification system that enables real-time

discrimination of sEMG segments based on the

extracted features and testing this system in human-

robot collaborative environments.

Lastly, it's important to note that while this study

primarily focuses on physiological signals, there are

also many human-robot interaction solutions based

on the analysis of non-bioelectrical data such as

kinematic or visual information. Although these

signals may not provide as much insight into the

neural processes underlying human behavior, they

offer benefits like easier recording and higher

accuracy in determining motion start and end points.

In general, the existence of such a variety of

approaches is a positive aspect within the scientific

community, and final integrative solutions will likely

emerge from a combination of these diverse

approaches.

ACKNOWLEDGEMENTS

The authors would like to thank the National Institute

of Advanced Industrial Science and Technology

(AIST) and the Japanese government for the

additional financial support provided through

KAKENHI grants (grant number: 18K18431).

REFERENCES

Abdi, H., "Holm’s sequential Bonferroni procedure," in

Encyclopedia of research design vol. 1(8), 2010, pp. 1-

Ahmed, S.N., "Covid, AI, and robotics—A neurologist's

perspective," in Frontiers in Robotics and AI, vol 8,

2021, pp. 617426.

Bacek, T., Moltedo, M., Langlois, K., Prieto, G.A.,

Sanchez-Villamañan, M.C., Gonzalez-Vargas, J.,

Vanderborght, B., Lefeber, D. and Moreno, J.C.,

"BioMot exoskeleton—Towards a smart wearable

robot for symbiotic human-robot interaction," in

International Conference on Rehabilitation Robotics

(ICORR), 2017.

Bainbridge, W.A., Nozawa, S., Ueda, R., Okada, K. and

Inaba, M., "A methodological outline and utility

assessment of sensor-based biosignal measurement in

human-robot interaction," in International Journal of

Social Robotics, vol. 4(3), 2012, pp. 303-316.

Bakal, D.A. and Kaganov, J.A., "Muscle contraction and

migraine headache: psychophysiologic comparison," in

Headache: The Journal of Head and Face Pain, vol.

17(5), 1977, pp. 208-215.

Characterization of sEMG Spectral Properties During Lower Limb Muscle Activation

711

Besley, T., and Case A., "Modeling technology adoption in

developing countries," in The American economic

review, vol. 83 (2), 1993, pp. 396-402.

Costa, Á., Iáñez, E., Úbeda, A., Hortal, E., Del-Ama, A.J.,

Gil-Agudo, A. and Azorín, J.M., "Decoding the

attentional demands of gait through EEG gamma band

features," in PLoS one, vol. 11(4), 2016, pp. e0154136.

Costa-García, Á., Iáñez, E., Sonoo, M., Okajima, S.,

Yamasaki, H., Ueda, S. and Shimoda, S.,

"Segmentation and averaging of sEMG muscle

activations prior to synergy extraction," in IEEE

Robotics and Automation Letters vol. 5(2), 2020, pp.

3106-3112.

Costa‐García, Á., Iáñez, E., Yokoyama, M., Ueda, S.,

Okajima, S. and Shimoda, S., "Quantification of high

and low sEMG spectral components during sustained

isometric contraction," in Physiological Reports vol.

10(10), 2022, pp. e15296.

Costa-García, Á., Úbeda, A. and Shimoda, S., "Effects of

Force Modulation on Large Muscles during Human

Cycling." in Brain Sciences vol. 11(11), 2021, pp. 1537.

Fratini, A., Cesarelli, M., Bifulco, P. and Romano, M.,

"Relevance of motion artifact in electromyography

recordings during vibration treatment," in Journal of

Electromyography and Kinesiology, vol. 19(4), 2009,

pp. 710-718.

Green, S. A., Billinghurst, M., Chen, X., and Chase, J. G,

"Human-robot collaboration: A literature review and

augmented reality approach in design," in International

journal of advanced robotic systems, vol. 5(1), 2008,

pp. 1.

Li, K., Zhang, J., Wang, L., Zhang, M., Li, J. and Bao, S.,

"A review of the key technologies for sEMG-based

human-robot interaction systems," in Biomedical

Signal Processing and Control, vol 62, 2020, pp.

102074.

Lienhard, K., Cabasson, A., Meste, O. and Colson, S.S.,

"sEMG during whole-body vibration contains motion

artifacts and reflex activity," in Journal of sports science

& medicine, vol. 14(1), 2015, pp. 54.

Lünenburger, L., Colombo, G. and Riener, R.,

"Biofeedback for robotic gait rehabilitation," in Journal

of neuroengineering and rehabilitation, vol. 4(1), 2007,

pp. 1-11.

Neblett, R., "Surface electromyographic (SEMG)

biofeedback for chronic low back pain," in Healthcare,

vol. 4(2), 2016, MDPI.

Rey, D. and Neuhäuser, M., "Wilcoxon-signed-rank test,"

in International encyclopedia of statistical science,

2011, pp. 1658-1659.

Ritchie, H., and Roser M., "Technology adoption," in Our

World in Data, 2017.

Robinson, E. and Clark, D., "Sampling and the Nyquist

frequency," in The Leading Edge vol. 10(3), 1991, pp.

51-53.

Shimoda, S., Jamone, L., Ognibene, D., Nagai, T., Sciutti,

A., Costa-Garcia, A., Oseki, Y. and Taniguchi, T.,

"What is the role of the next generation of cognitive

robotics?," in Advanced Robotics, vol. 36(1-2), 2022,

pp. 3-16.

Stegeman, D. and Hermens, H.,"Standards for surface

electromyography: The European project Surface EMG

for noninvasive assessment of muscles (SENIAM)," in

Enschede: Roessingh Research and Development, vol.

10, 2007, pp. 8-12.

Xi, X., Tang, M., Miran, S.M. and Luo, Z., "Evaluation of

feature extraction and recognition for activity

monitoring and fall detection based on wearable sEMG

sensors," in Sensors, vol, 17(6), 2017, pp. 1229.

Yang, J.C., Mun, J., Kwon, S.Y., Park, S., Bao, Z. and Park,

S., "Electronic skin: recent progress and future

prospects for skin-attachable devices for health

monitoring, robotics, and prosthetics," in Advanced

Materials, vol 31(48), 2019, pp. 1904765.

BIOSIGNALS 2024 - 17th International Conference on Bio-inspired Systems and Signal Processing

712