Development of Myoelectric Control Module for Prosthetic Hand

with Artifact Removal during Sensory Electrical Stimulation

Yashuo Yu

, Chih-Hong Chou

1,2 b

, Jie Zhang

, Manzhao Hao

1,2 d

and Ning Lan

1,2 e

School of Biomedical Engineering, Shanghai Jiao Tong University, Shanghai, China

Institute of Medical Robotics, Shanghai Jiao Tong University, Shanghai, China

Keywords: Prosthetic Hand, Sensory Feedback, Transcutaneous Electrical Nerve Stimulation (Tens), Stimulation Artifact.

Abstract: Evoked Tactile Sensation (ETS) with transcutaneous electrical nerve stimulation (TENS) can provide

amputees with a non-invasive neural interface for sensory feedback. However, sensory stimulation at the

projected finger map (PFM) on the stump skin causes interference in surface electromyographic (sEMG)

signals used for prosthesis control. This study developed a practical solution that combined hardware blanking

and software filtering to eliminate stimulus artifacts in real-time. A synchronized blanking circuit was inserted

after the differential amplifiers to partially remove artifact spikes. EMG signal was then sampled and further

processed by a digital signal processor (DSP). A digital comb filter removed the remaining artifacts at all

harmonic frequencies of stimulation. The filtered EMG was rectified, and its envelope was extracted to control

prosthetic hand. This technique was tested for its effectiveness in removing stimulus artifacts in three able-

bodied subjects and in one transradial amputee operating a Bebionic hand. Results in able-bodied subjects

indicated that the technique was effective in removing stimulus artifacts in EMG under different conditions.

In the amputee subject, grasp control using the Bebionic hand was obtained with simultaneous sensory

stimulation in the ipsilateral stump. The amputee subject achieved an average success rate of 90% for

identifying the length of grasped objects. Tests confirmed that the technique is adequate to remove stimulus

artifacts from EMG signals and allows control of the Bebionic hand with simultaneous sensory stimulation.

1 INTRODUCTION

Commercial prosthetic hands employ surface

electromyographic (sEMG) signals from residual

muscles for motor control. However, a survey of

amputees reported that the low prevalence and high

abandonment rate of such devices are attributed to the

lack of sensory function (Smail et al., 2021), which

plays an important role in the daily activities (e.g.

grasping tasks) for able-bodied subjects (Johansson &

Flanagan, 2009). Although there have been many

studies on closed-loop prosthetic systems in recent

years, the progress has been limited in the laboratory

(Bensmaia et al., 2020). Therefore, the restoration of

sensory function of a myoelectric-controlled bionic

hand has been a great challenge in neural engineering.

https://orcid.org/0000-0003-3512-4954

https://orcid.org/0000-0002-8985-5050

https://orcid.org/0000-0002-6075-2074

https://orcid.org/0000-0001-8744-7128

https://orcid.org/0000-0001-6061-5419

Our previous studies established that sensory

feedback delivered with transcutaneous electrical

nerve stimulation (TENS) to the projected finger map

(PFM) of the amputated stump can generate evoked

tactile sensation (ETS), and revealed that PFM on the

stump skin of the forearm of amputees corresponded

to the projection of the whole hand (Chai et al., 2015).

This afferent information shares a natural and direct

pathway with intact tactile afferents to the primary

somatosensory cortex (SI) (Hao et al., 2020). The

finger-specific perceptibility affords the advantage of

multi-facet recognition during prosthetic grasping (Li

et al., 2021).

In a closed-loop myoelectric prosthesis with non-

invasive sensory stimulation by TENS, it also

requires to collect sEMG signals for motor control,

118

Yu, Y., Chou, C., Zhang, J., Hao, M. and Lan, N.

Development of Myoelectric Control Module for Prosthetic Hand with Artifact Removal during Sensory Electrical Stimulation.

DOI: 10.5220/0010778600003123

In Proceedings of the 15th International Joint Conference on Biomedical Engineering Systems and Technologies (BIOSTEC 2022) - Volume 1: BIODEVICES, pages 118-125

ISBN: 978-989-758-552-4; ISSN: 2184-4305

Figure 1: System block diagram of the integrated myoelectrical control and ETS-based sensory feedback of a Bebionic

prosthetic hand.

where the placement location of sEMG sensors

and stimulation electrodes are very close to each other

at the stump ipsilaterally. Stimulation artifacts

conducted through the skin and muscles can produce

a stimulus artifact much larger than EMG signals

from voluntary contractions. Accordingly, a major

drawback of TENS applied to the stump skin is that it

generates interference to sEMG signals.

A simple way to handle EMG contaminated with

stimulus artifacts from hardware is to wait for the

recovery of EMG amplifiers from saturation, which

requires a fast recovery of electronic components or

to switch off the amplifier (mute) when the stimulus

pulse starts to be delivered and discard the data during

saturation (Schauer et al., 2004). Besides, sample and

hold circuit can also be used for removal by storing

the DC levels before the pulse and holding it during

the pulse (Babb et al., 1978).

Software-based approaches combining

embedded processors allow for real-time or semi-

real-time removal. Software algorithms for real-time

processing need not only to achieve further

enhancements than hardware processing but also to

ensure instantaneity, which requires appropriate

complexity. However, most of algorithms for post-

processing focus on improving the algorithm

accuracy instead of efficiency. The software methods

for suppressing EMG signal artifacts can be broadly

classified into the following categories according to

the existing research: 1) blanking (Yi et al., 2013; Li

et al., 2019), 2) comb filter (Frigo et al., 2000;

Widjaja et al., 2009), 3) adaptive filter (Qiu et al.,

2015), and 4) other methods (Yochum et al., 2014;

Pilkar et al., 2016; Bi et al., 2021).

Dosen et al. (Dosen et al., 2014) used a time-

division multiplexing (TDM) approach with similar

logic to sample and hold, performing stimulation and

recording in dedicated, non-overlapping time

windows to avoid interference, and thus building a

real-time closed-loop control system.

The various solutions of artifacts removal are not

readily available for practical use in closed-loop

control prosthetic systems. In this study, we

developed and tested a module composed of a hybrid

method of hardware blanking and software filtering

to remove stimulus artifacts in sEMG caused by

TENS. This technique can allow integration of

control of myoelectrical prosthesis with the ETS-

based sensory feedback in the ipsilateral stump of

amputees. We verified this method with a

commercially available Bebionic hand to form a

closed-loop system. Test results indicated that the

prosthesis works properly with simultaneous sensory

perception of grasp, demonstrating that sensory

feedback adds the functionality of the device. The

system block diagram is shown in Figure 1. The blue

line represents control pathway, and the red line

represents sensory feedback pathway.

2 METHODS

2.1 Module Overview

The framework of the myoelectric control module for

prosthetic hand with artifact removal is shown in the

block diagram in Figure 2. In the control pathway, the

EMG signals from residual muscles are collected with

surface electrodes. The signals are blanked (switch

off) by a hardware circuit during stimulation and

filtered to remove the remaining artifacts. The

processed EMG signals are rectified, and the

envelopes are used to control the opening and closing

of prosthetic hand. EMG sampling, artifact filtering,

rectifying and envelope extraction are all processed in

Development of Myoelectric Control Module for Prosthetic Hand with Artifact Removal during Sensory Electrical Stimulation

119

Figure 2: The hardware block diagram of the myoelectric control module.

a digital signal processor (DSP). In the sensory

pathway, contact pressures at prosthetic fingertips are

converted to a pattern of electrical stimuli and

delivered to five stimulation electrodes via the multi-

channel stimulator, which can generate specific

stimulation patterns according to an encoding

protocol. The stimulation patterns are charge-

balanced biphasic rectangular pulse trains with

positive pulses going first. Sensory intensity is

modulated by pulse amplitude (PA), pulse frequency

(PF) and pulse width (PW). Changes in stimuli

parameters can cause different sensory modalities, as

in (Yang et al., 2020).

2.2 System Description

To reduce the effect of the electrical stimulus on the

signal acquisition, we designed a blanking circuit to

turn off the electrical stimulus input to avoid the

current saturating the capacitor. We also moved the

switch to the output of the instrumentation amplifier

(OPA2333, Texas Instruments, USA) for reducing

the signal noise during switching (Rolston et al.,

2009). The architecture of the system is shown in

Figure 2.

When grasping occurs, the sensor in the prosthetic

hand generated the electrical stimulation signal. The

output of the stimulation sent the electrical

stimulation trigger to the blanking circuit

simultaneously. Then, the system turned off the meter

amplifier output within 10 μs, and conversely turned

on the acquisition function within 10 μs after the

electrical stimulation ends.

The subsequent part is a typical EMG signal

processing circuit. After band-pass filtering and A/D

conversion, the signal was processed with the comb

filter (refer to details in Section 2.3) implemented by

DSP. Then, the EMG envelope features were

extracted. Finally, the signal through co-active

mapping of one pair of antagonistic muscles was used

as the output control of the prosthesis. In addition, a

dead zone was set to avoid driving prosthesis under

weak EMG, which may lead to frequent jitter.

2.3 Signal Processing

After being propagated as local currents, stimulation

pulses manifest in the frequency domain as a

fundamental wave of the pulse delivery frequency

and its higher harmonics, which can be removed with

a comb filter with the following equation,

𝑦

(

𝑛

)

𝑥

(

𝑛

)

−𝑥

(

𝑛−𝑁



)

√

(1)

where x(n) and y(n) are for the raw and the filtered

EMG at sampling time n. N

Tstim

is the number of

samples between inter-pulse-intervals (IPIs). The

scale factor √2 can keep the signal power constant

after filtering (Frigo et al., 2000).

Then EMG signals were then filtered through a

6th Butterworth lowpass filter with a 400 Hz cutoff

frequency to reject high frequency components.

Finally, the signals were rectified and filtered with

2nd lowpass Butterworth with 10 Hz cutoff to extract

envelope curves.

All filters were implemented on a DSP

(STM32H743, STMicroelectronics, Italy) to provide

real-time operation.

3 EXPERIMENTS

3.1 Subjects

Three able-bodied subjects (2 males and 1 female,

29±6.083 yrs.) and 1 transradial amputated subject

(male, 54 years old, left amputation) were recruited to

BIODEVICES 2022 - 15th International Conference on Biomedical Electronics and Devices

120

participate in this study. All the experiment protocols

were approved by the Institutional Review Board for

Human Research Protections, Shanghai Jiao Tong

University.

3.2 Experiment Setup

The subject was seated at a table, and a Bebionic hand

(Otto Bock HealthCare) was fixed with an upright

tripod. Two pairs of bipolar EMG electrodes

(Norotrode 20 Bipolar sEMG Electrodes) were

placed on ulnar/radial side of forearm near elbow end

to capture two-channel EMG signals of wrist

flexor/extensor. The prosthetic hand was set to close

when the flexor contracted and to open with extensor

contraction (Figure 3 (a)). A non-invasive stimulation

pattern generator (Liu et al., 2015), delivered five

channels of TENS to five metal stimulation electrodes

on the forearm skin (Figure 3 (b), (c)) for able-bodied

subjects or the stump PFM site (for amputees)

corresponding to pressures (meansured by force

sensor, FlexiForce A201, Tekscan, Inc., USA) on

each digit tip (Figure 3 (d)). The stimulation electrode

was a disk Ag/AgCl electrode with 10 mm diameter,

custom made by the Institute of Semiconductors of

the Chinese Academy of Sciences. The reference

electrode (non-woven fabric circular electrode with 5

cm diameter) of each channel was placed near the

olecranon. The system equipped with EMG

acquisition function can output two channels of co-

active mapped execution signals to manipulate the

prosthesis. Both stimulation and EMG processing

were controlled by a host PC with self-innovate

software designed by C#. The two-channel EMG

signals were sampled at 2 kHz.

3.3 Experiment Protocols

There were two experiments in total. In the first

experiment, the able-bodied subjects explored

whether artifact removal strategies were effective

under different settings, and the second experiment

was carried out to verify the ETS-based sensory

feedback in real-time for closed-loop control on

amputees, which can be seen as a variant of (Li et al.,

2021) but in ipsilateral.

The 50 Hz fixed stimulation frequency was used

throughout the experiment, while pulse amplitude

(PA) and pulse width (PW) were modulated to encode

prosthetic hand contact pressure (Yang et al., 2020).

Figure 3: (a) Experiment scene. (b) and (c) show far and

near electrode distances respectively on able-bodied

subjects while (d) shows the amputee’s electrode sites with

stimulation electrodes on PFM. (e) 4 blocks of different

lengths (VS, S, M, and L) used in experiment 2, allowing

contact of different numbers of fingers. Fingers involved in

pinch are, from left to right: thumb and index thumb, index

and middle; thumb, index, middle and ring; thumb, index,

middle, ring and little.

3.4 Experiment 1: Filtering Algorithm

Validity Test

There were two conditions of stimulation electrode

placement: 1) far condition: stimulation electrodes

were placed between one half and three quarters

(distal end) of the full length (measured by the

distance from olecranon to ulnar styloid process) of

the forearm far from the EMG electrodes (Figure 3

(b)); 2) near condition: between two EMG electrodes

Development of Myoelectric Control Module for Prosthetic Hand with Artifact Removal during Sensory Electrical Stimulation

121

(Figure 3 (c)). The former condition was determined

according to the actual length of amputees’ residual

limb. The stimulation intensity level was set with a

PW of 600μs per channel. Then the PA of each

channel was set to the upper limit value of buzz

sensation multiplied by 0.8 when 5 channels

delivering simultaneously to avoid

uncomfortableness.

The experimental task was designed to meet the

practical application of the prosthetic for hand

opening and closing movement. The subjects

recorded their EMG by flexing/extending their wrist

for about 2s with stimulation. Ten trials for each

combination of electrode distances and

flexion/extension, that was 40 trials in total for each

subject.

Power generated by stimulus artifacts was further

quantified. First, EMG of able-bodied subjects was

segmented, and each segment contained 4000 sample

points (2 s signal sampled at 2 kHz). Then the power

spectral densities (PSD) were calculated by Welch’s

method. Eventually the PSD at each artifact harmonic

with 10Hz on each side was accumulated to estimate

the artifact power (P

stim

) with Eq. (2),

𝑃

𝑠𝑡𝑖𝑚

= 𝑃𝑆𝐷

(

𝑖

)

𝑘∗𝑓

𝑠𝑡𝑖𝑚

+10

𝑖=𝑘∗𝑓

𝑠𝑡𝑖𝑚

−10

𝑓𝑠/(2∗𝑓

𝑠𝑡𝑖𝑚

)

𝑘=1

(2)

where k is the number of harmonics and i is the

specific value of the discrete spectrum.

The averaged P

stim

in every condition over all

subjects was calculated. And the electrode distance

was considered as an intra-class factor to compare the

mean P

stim

, whose significant differences were tested

with two-tailed paired Tukey’s honestly significant

difference test.

3.5 Experiment 2: System Operation

Validation Test

In experiment 2, a response button was placed on

amputated subject’s contralateral hand. The

experimental task was to bend the wrist so that the

prosthetic hand can touch the wooden block when

closing. There were 4 different lengths of the blocks,

which could be represented by VS (very small), S

(small), M (medium), and L (large) respectively.

Hence, the number of fingers varied when the hand

grasped the block (Figure 3 (e)). During the contact

process between the prosthetic hand and the wood

block, the subject needed to determine the block size

by the finger numbers feeling through ETS. 5

channels of pressure signals, prosthetic hand aperture

(distance between tips of thumb and index), and the

response signal was sampled at 100 Hz synchronized

with EMG. Each grasping and identification make a

single trial and the experiment contained 40 trials in

total, with 10 trials for each size pseudo-random

ordered.

4 RESULTS

4.1 Effects of Artifact Removal at

Different Stages

The representative EMG signals at each processing

stage are plotted in Figure 4. Raw EMG without any

blanking is contaminated by a large amplitude of

stimulus artifacts with 1 channel of stimulation at 1

mA 200 μs. The artifacts do not saturate the

differential amplifiers. It is clear that artifacts are

significantly diminished with hardware blanking. It is

noted, however, that recording electrodes place

nearer to the stimulation electrodes yield fewer

artifacts than those with greater distance. Moreover,

contaminated EMG signals display artifact

components in all harmonic frequencies. After

processing by the comb and Butterworth filters,

artifacts of all harmonic components are largely

invisible in the filtered EMG. In spite of some loss of

EMG components at the harmonic frequencies, the

envelope of processed EMG signals can still be

extracted.

4.2 Filtering Algorithm Validity

Results

Figure 5 summarizes results across different

conditions in three able-bodied subjects. The vertical

axis shows spectral power in dB. Statistic analysis

indicates that P

stim

decreased along artifacts removal

stages regardless of muscle and the distance between

EMG and stimulation electrodes. It is clear that each

stage of hardware and software processing of this

method results in a certain degree of removal of

artifact components from the contaminated EMG. It

is expected that, artifacts power was the largest with

no blanking; and artifacts power is the smallest with

blanking and digital filtering. Artifacts are stronger

for EMG recording electrodes placed farther away

from the stimulation electrodes.

BIODEVICES 2022 - 15th International Conference on Biomedical Electronics and Devices

122

Figure 4: Extensor EMG signal and corresponding one-sided FFT spectrums at each processing stage. From top to bottom,

the respective stages are raw EMG without blanking, EMG with hardware blanking, EMG with software filtering, and the

final envelopes. The left and right columns are two cases where distance between stimulation and recording electrodes is near

and far with each other, respectively.

Figure 5: Bar plot of the harmonic power P

stim

(mean ±

standard deviation) shows the average results of 3 able-

bodied subjects across conditions. (Asterisks represent the

statistically significant difference between the near and far

intra-class conditions. **, p < 0.01; ***, p < 0.001).

Besides, all subjects show significant differences between

different stages and different electrode placement

conditions, which is not shown in the figure.

4.3 System Operation Validation

Results

Figure 6 shows signals from both feedforward control

and sensory feedback of the closed-loop prosthetic

system.

It is shown that the difference between EMG

envelops of the flexor and the extensor increases

initially from zero to a positive value for hand

closing, then jitters around dead zone, later decreases

to a negative value for hand opening, and finally

returns to zero at resting hand posture. This implies

that control signals sent to the prosthetic hand first

close the hand for grasping the object and then opens

to release the object. This can also be seen from the

plots of hand apertures at the bottom of Figure 6. It

can be marked by several phases/timings (t0~t5). The

t0 is the beginning of trial (waiting state). The t1

stands for the start of flexor recruitment by amputee

for prosthesis operation. The EMG envelope

difference rapidly increases as the prosthesis begins

to close. The t2 stands for the timing that prosthetic

fingers contact with the block, at which the prosthetic

hand stops at the closing position, and the pressure

starts to rise. The red arrow indicates the size identify

response from the subject through ETS, and the

period between t3 and t1 represents identification

time. At the t4, extensor contraction starts. However,

the amplitude of EMG envelope difference between

t4 and t5 fails to reach the level to drive the prosthesis,

thus forming a dead zone. Then at hand opening stage

t5, the difference reverses to a sufficient value below

0. Finally, when the hand is fully open at t6, contact

pressure also drops along with the process.

Development of Myoelectric Control Module for Prosthetic Hand with Artifact Removal during Sensory Electrical Stimulation

123

During the hand closing phase, in the feedforward

pathway the envelope difference is positive,

corresponding to the subsequent rising in contact

pressure and sensory feedback and vice versa. The

correlations among these signals illustrate that the

closed-loop system is functioning as expected in real-

time in the grasping task.

Figure 7 shows the confusion matrix of size

identification by the amputee subject. The accuracy

of identification for VS, S, M, and L sizes are 100%,

90%, 80%, 90% respectively. It can be seen that in

general the rate of misjudgement increased with the

number of contacting digits. However, M size is the

most likely one to be misjudged. This result confirms

that grasping and perception can be performed

ipsilaterally and simultaneously with the closed-loop

control system.

Figure 6: The illustration of the output signals by amputee

and prosthesis phases during operation of the closed-loop

control system. The bottom illustrates the change in

prosthesis operation (hand apertures).

Figure 7: Confusion matrix presents the accuracy of size

identification.

5 DISCUSSION AND

CONCLUSION

This study applied a combination of hardware and

software processing for artifact removal from TENS.

The myoelectrical prosthetic hand can be integrated

with the ETS-based sensory feedback in real-time for

closed-loop control on ipsilateral side for amputees

without data segment or artifact template database

establishment. This enabled the user to operate the

prosthetic hand to grasp objects and obtain sensory

perception from specific fingers.

In order to remove stimulus artifacts caused by

TENS with a relatively large amplitude of stimulation

current, hardware circuitry of the EMG acquisition

was modified to blank the output of EMG differential

amplifiers in real-time triggered by stimulation pulses

at each pulse delivery. Test indicated that this method

eliminated part of stimulus artifacts, but allowed for

maximum retention of EMG signals. In addition,

digital filtering algorithms were adopted to remove

remaining artifacts using a DSP. The difference of the

envelopes of processed EMG signals was shown to be

adequate as instructions to control the commercial

prosthetic hand (Figure 6).

Results in Figure 5 indicate that stimulation

electrodes placed farther away from the recording

electrodes generated larger artifacts in the experiment

of this study. This may be due to the fact that stimulus

current trans-passed the recording electrodes. Thus,

the longer the distance between stimulation and

recording electrodes, the larger the volume resistance

between them, yielding a larger stimulus artifact. In

future applications, the reference electrodes for

stimulation may be relocated to other places to further

reduce the effect of stimulus artifacts.

Our study mainly shows that the receiving of

sensory feedback and the control of prostheses can be

integrated on ipsilateral side for amputees during

TENS pulses delivering. The technique can be

adequate to allow integrating myoelectric control of a

commercial prosthetic hand with ETS-based sensory

feedback, and can be further applied to other closed-

loop systems with interference between pathways.

There is still room to improve signal processing

methods. Other types of digital filters, such as

Bayesian filters, could be employed for artifacts

generated by higher stimulation frequencies. In

addition, the amplitude of EMG envelopes varies

with subjects and placement of EMG electrodes.

Thus, in future applications, parameters of envelope

gain should be adjustable according to each user for

optimal closed-loop operation of control and sensory

perception.

BIODEVICES 2022 - 15th International Conference on Biomedical Electronics and Devices

124

ACKNOWLEDGEMENTS

This research was supported in part by a grant from

Key-Area Research and Development Program of

Guangdong Province (2020B0909020004), the

National Key R&D Program of China (No.

2017YFA0701104, No. 2020YFC2007903), a grant

from the National Natural Science Foundation of

China (No. 81630050), and a grant from Science and

Technology Commission of Shanghai Municipality

(No. 20DZ2220400).

REFERENCES

Babb, T. L., Mariani, E., Strain, G. M., Lieb, J. P., Soper,

H. V., & Crandall, P. H. (1978). A sample and hold

amplifier system for stimulus artifact suppression.

Electroencephalography and Clinical

Neurophysiology, 44(4), 528–531.

Bensmaia, S. J., Tyler, D. J., & Micera, S. (2020).

Restoration of sensory information via bionic hands.

Nature Biomedical Engineering, 1–13.

Bi, Z.-Y., Zhou, Y.-X., Xie, C.-X., Wang, H.-P., Wang, H.-

X., Wang, B.-L., Huang, J., Lü, X.-Y., & Wang, Z.-G.

(2021). A hybrid method for real-time stimulation

artefact removal during functional electrical stimulation

with time-variant parameters. Journal of Neural

Engineering, 18(4), 046028.

Chai, G., Sui, X., Li, S., He, L., & Lan, N. (2015).

Characterization of evoked tactile sensation in forearm

amputees with transcutaneous electrical nerve

stimulation. Journal of Neural Engineering, 12(6),

066002.

Dosen, S., Schaeffer, M.-C., & Farina, D. (2014). Time-

division multiplexing for myoelectric closed-loop

control using electrotactile feedback. Journal of

NeuroEngineering and Rehabilitation, 11.

Frigo, C., Ferrarin, M., Frasson, W., Pavan, E., & Thorsen,

R. (2000). EMG signals detection and processing for

on-line control of functional electrical stimulation.

Journal of Electromyography and Kinesiology, 10(5),

351–360.

Hao, M., Chou, C.-H., Zhang, J., Yang, F., Cao, C., Yin, P.,

Liang, W., Niu, C. M., & Lan, N. (2020). Restoring

Finger-Specific Sensory Feedback for Transradial

Amputees via Non-Invasive Evoked Tactile Sensation.

IEEE Open Journal of Engineering in Medicine and

Biology, 1, 98–107.

Johansson, R. S., & Flanagan, J. R. (2009). Coding and use

of tactile signals from the fingertips in object

manipulation tasks. Nature Reviews Neuroscience,

10(5), 345–359.

Li, Y., Chen, J., & Yang, Y. (2019). A Method for

Suppressing Electrical Stimulation Artifacts from

Electromyography. International Journal of Neural

Systems, 29(06), 1850054.

Li, Y., Chou, C.-H., Zhang, J., Zhang, Z., Hao, M., & Lan,

N. (2021). A Pilot Study of Multi-Site Simultaneous

Stimulation for Tactile and Opening Information

Feedback in the Prosthetic Hand. 2021 10th

International IEEE/EMBS Conference on Neural

Engineering (NER), 187–190.

Liu, X. X., Chai, G. H., Qu, H. E., & Lan, N. (2015). A

sensory feedback system for prosthetic hand based on

evoked tactile sensation. 2015 37th Annual

International Conference of the IEEE Engineering in

Medicine and Biology Society (EMBC), 2493–2496.

Pilkar, R., Ramanujam, A., Garbarini, E., & Forrest, G.

(2016). Validation of empirical mode decomposition

combined with notch filtering to extract electrical

stimulation artifact from surface electromyograms

during functional electrical stimulation. 2016 38th

Annual International Conference of the IEEE

Engineering in Medicine and Biology Society (EMBC),

1733–1736.

Qiu, S., Feng, J., Xu, R., Xu, J., Wang, K., He, F., Qi, H.,

Zhao, X., Zhou, P., Zhang, L., & Ming, D. (2015). A

Stimulus Artifact Removal Technique for SEMG

Signal Processing During Functional Electrical

Stimulation. IEEE Transactions on Biomedical

Engineering, 62(8), 1959–1968.

Rolston, J. D., Gross, R. E., & Potter, S. M. (2009). A Low-

Cost Multielectrode System for Data Acquisition

Enabling Real-Time Closed-Loop Processing with

Rapid Recovery from Stimulation Artifacts. Frontiers

in Neuroengineering, 2.

Schauer, T., Salbert, R. C., Negard, N.-O., & Raisch, J.

(2004). Detection and Filtering of EMG for Assessing

Voluntary Muscle Activity during FES. 185–187.

Smail, L. C., Neal, C., Wilkins, C., & Packham, T. L.

(2021). Comfort and function remain key factors in

upper limb prosthetic abandonment: Findings of a

scoping review. Disability and Rehabilitation: Assistive

Technology, 16(8), 821–830.

Widjaja, F., Shee, C. Y., Poignet, P., & Ang, W. T. (2009).

FES artifact suppression for real-time tremor

compensation. 2009 IEEE International Conference on

Rehabilitation Robotics, 53–58.

Yang, F., Hao, M.-Z., Zhang, J., Chou, C.-H., & Lan, N.

(2020). An Experimental Protocol for Evaluating Pulse

Width Modulation Ranges of Evoked Tactile Sensory

Feedback in Amputees*. 2020 42nd Annual

International Conference of the IEEE Engineering in

Medicine Biology Society (EMBC), 3869–3872.

Yi, X., Jia, J., Deng, S., Shen, S. G., Xie, Q., & Wang, G.

(2013). A Blink Restoration System With Contralateral

EMG Triggered Stimulation and Real-Time Artifact

Blanking. IEEE Transactions on Biomedical Circuits

and Systems, 7(2), 140–148.

Yochum, M., Bakir, T., Binczak, S., & Lepers, R. (2014).

EMG artifacts removal during electrical stimulation, a

CWT based technique. 2014 IEEE Region 10

Symposium, 137–140.

Development of Myoelectric Control Module for Prosthetic Hand with Artifact Removal during Sensory Electrical Stimulation

125