The Approach to the Detection of the Movement Precursor by

Electromyographic Signals

Semen Kurkin

1 a

, Vladimir Khorev

2 b

, Elena Pitsik

1 c

, Vladimir Maksimenko

1 d

and

Alexander Hramov

1 e

Neuroscience and Cognitive Technology Lab, Innopolis University, Innopolis, Russian Federation

Department of Dynamical Modeling and BioMedical Engineering, Saratov State University, Astrakhanskaya Street 83,

Saratov, Russian Federation

Keywords: EMG, Electromyogram, Movement Detection, Data Analysis, Pattern Recognition.

Abstract: We have developed a technique allowing automatic detection of the precursor of movement beginning based

on the analysis of electromyographic signals. Methods for determining the beginning of movement and the

moments of movement planning are of urgent need in neuroscience, and a separate problem is the use of

muscle electrical activity signals (electromyograms) to accurately determine the beginning of hand movement

due to the complexity, short duration and noise of the original signals. This issue is particularly significant

for experiments with simultaneous recording of electroencephalograms, when it is necessary to consider the

interaction between the structures of the brain. We have found out that in the case when the movement starts

on a certain sound signal, the moment of the movement beginning is detected with a some time delay.

1 INTRODUCTION

The development of effective methods for

determining the precursor of movement beginning

and the moments of movement planning is an urgent

problem of neuroscience and neurotechnology. In

particular, this task is closely related to the

development of human-machine interfaces. A

separate problem here is the use of muscle electrical

activity signals (electromyograms) for exact

detection of the of limb movement precursor. This

issue is particularly acute when conducting

experiments with simultaneous recording of

electroencephalograms, when the relationship

between the excitation of certain brain areas and

human motor activity is investigated.

Currently, the problem of studying the processes

occurring in human body related to the performance

of motor activity attracts a large scientific interest

(Wood et al., 2014; Hayashibe et al., 2015;

Maksimenko et al, 2018; Mondini et al., 2018). The

https://orcid.org/0000-0002-3438-5717

https://orcid.org/0000-0001-6613-8940

https://orcid.org/0000-0003-1850-2394

https://orcid.org/0000-0002-4632-6896

https://orcid.org/0000-0003-2787-2530

relevance of this research area is connected with the

possibility of applying the results in such areas as

rehabilitation, prosthetics, robotics and others.

However, the use of additional methods for the

analysis of motor activity, basically, involves

conducting an experiment according to a previously

developed plan, according to which movements are

performed on a special sound signal. In this case,

there is the problem of accurate determining the

moment of the start of movement (Reis, 2014).

To solve this problem, it seems promising to use

signals of electrical activity excited directly by

muscle fibers – electromyograms (EMG) (Rouillard

et al., 2015). The analysis of EMG signals, in turn, is

difficult due to the low amplitude of the potentials,

the strong nonstationarity, the presence of various

artifacts and the poor structuring of the initial data

(Basmajian, 1979; De Luka, 2010; Kastalskiy et al.,

2018).

Thus, there is now a need to develop new effective

methods for EMG signals analysis and for their

276

Kurkin, S., Khorev, V., Pitsik, E., Maksimenko, V. and Hramov, A.

The Approach to the Detection of the Movement Precursor by Electromyographic Signals.

DOI: 10.5220/0007916502760280

In Proceedings of the 16th International Conference on Informatics in Control, Automation and Robotics (ICINCO 2019), pages 276-280

ISBN: 978-989-758-380-3

application for a detailed study of human motor

activity.

2 METHODS

2.1 Data Preparation

In the course of the experiment, the registration of

non-invasive EMG signals from the elbow muscle

was carried out. The subject was in an upright

position. The subject had no pathologies of the central

nervous system.

The duration of the experiment was 150 minutes.

During the recording of signals, breathing was

arbitrary.

The subject was instructed to perform on the

sound signal the following actions: (1) flexion and (2)

subsequent extension of the hand with intermediate

fixation in the upper position (see Fig. 1a).

Registration of EMG signals was carried out using a

multichannel electroencephalograph-analyzer

“Encephalan-131-03”, model 10 (Taganrog, Russian

Federation) with a set of standard sensors. Signals

were recorded at a sampling frequency of 250 Hz with

a 12-bit resolution.

For additional control and registration of motor

activity, a copy-type setting device was used, which

is a lever construction made of plastic and light

alloys, made similar to the human skeletal scheme

with the coincidence of the position of the mobility

axes and joints (exoskeleton). The lever mechanism

was identical to the kinematic scheme of a human

hand and contained an analogue of the forearm

connected to the shoulder with a rotational pair with

one degree of freedom, which allows to obtain data

on the flexion of the elbow joint simultaneously with

EMG recording.

2.2 Experimental Setup

The structure of the experiment is shown in Fig. 1c.

In total, the experiment consisted of six sessions and

included pre-registration of background activity

without subject performing special instructions (BG,

15 minutes), two half-hour sessions with flexion of

the hand on the sound signal (AM), two sessions with

arbitrary flexion of the hand (FM), the final

registration of the background activity without the

subject performing special instructions for 15

minutes. The beginning of each session was preceded

by an automatic audiovisual warning of the subject

about its occurrence. For sessions with flexion of the

hand on the sound signal 50 movement repetitions

were planned. Sound stimuli were given at arbitrary

moments of time but provided for at least 10 seconds

of rest between every two movements. For the session

with an arbitrary flexion of the hand, no sound stimuli

were given, however, the subject was instructed to be

at rest also for at least 10 seconds after each period of

motor activity. The experiment was conducted in the

first half of the day in a specially equipped laboratory,

where the volunteer was in comfortable environment,

eliminating the presence of distracting factors, such as

background noise and bright light.

2.3 Methods

To detect the precursor of the hand movement, the

EMG signal was numerically filtered in the frequency

band 1–10 Hz, then smoothed by a sliding window of

2 s in length, after which the derivative of the signal

was found along the smoothed series.

Figure 1: (a) A schematic representation of the movement of

the subject's hand with a connected sensor for measuring the

EMG signal during the experiment; 1 corresponds to the

extension of the hand, and 2 – to the flexion. (b) Photograph

of the subject's arm with sensors for measuring EMG signals

and exoskeleton. (c) The structure of the experiment that

contains the following sessions: BG, AM and FM denote a

single period of background activity, audio stimulated

movement and free movement, respectively. Each session

was preceded by the video message with instructions.

Comparing the original EMG signal and the derived

The Approach to the Detection of the Movement Precursor by Electromyographic Signals

277

derivative, it was found that at the time points

corresponding to the beginning of the movement, value

of the derivative exceeds the threshold value.

Thus, comparing the received signal with the

threshold value at each moment of time, the moments

corresponding to the beginnings of movement

(precursors of the movement) were determined.

Figure 2 shows a fragment of typical raw EMG signal

recorded from an elbow muscle (Fig. 2a), smoothed

time series of the EMG (Fig. 2b), and its derivative

(Fig. 2c). The grey line corresponds to the threshold

value of the derivative used for automatic detection

of the moment of movement beginning.

The red risks mark the moments of the sound signals

corresponding to the commands. The Fig. 2 shows

that a sharp increase in the amplitude of the registered

signal corresponds to the moments of the beginning

of the movements.

3 RESULTS

On the basis of the conducted research, the optimal

threshold value of the derivative of the EMG signal

was determined, which ensures the best ratio of the

sensitivity and the percentage of false conclusions of

the algorithm for determining the moments of the

movement beginning. This value is equal to 0.5 of the

maximum value of the series derivative.

Fig. 3 shows the dependencies of true positives (TP)

and false positives (FP) of the algorithm for

determining the movement beginning on the

threshold value R. The threshold value varied in the

range from zero to the maximum value of the series

with a step of 0.05. It is clearly seen from Fig. 3 that

the maximum difference between TP and FP (the best

accuracy of the algorithm) is observed at the value of

Figure 2: (a) Fragment of the original (raw) experimental EMG signal; (b) smoothed and filtered EMG signal (blue curve);

(c) its derivative (green curve). The moments of the sound signals are marked in red, the line of the threshold value used to

determine the moments of the beginning of the movements is marked with the grey line (threshold). (d) Enlarged fragment

of the filtered signal and its derivative, which demonstrates the delay h between the moment of presentation of the sound

signal and the moment of beginning of the movement.

ICINCO 2019 - 16th International Conference on Informatics in Control, Automation and Robotics

278

R equal to 0.5 of the maximum value of the series.

Note, that this value of R was then used to calculate

the distribution of delays h between the time of sound

signal presentation and the beginning of the

movement. The advantage of this approach is its

simplicity and speed, compared with more accurate

and complex methods that require individual training

of subjects.

Indeed, it can be seen from Fig. 2 that there is a time

delay h between the moment when the sound signal is

presented, and the detected time moment (precursor)

of movement beginning. The analysis of the

characteristic time delay h is shown in Fig. 4 in the

form of the distribution obtained in one typical

experimental session. The blue curve in the figure is

Gaussian kernel density estimate.

Figure 3: Dependencies of true positives (TP, red curve)

and false positives (FP, blue curve) of determining the

movement beginning on the threshold value R. The

dependencies are given in percent.

It can be seen that the mode of the distribution

corresponds to the time 1.6–1.8 s. The narrowness of

the distribution obtained suggests that the preparation

time for movement can be estimated and then

considered in the experiment without presenting a

sound signal. The causes of the time delay detected in

the work may be related to the processes of the

stimulus processing and movement planning. In this

context, the use of EMG signals provides great

potential for identifying the various phases associated

with the implementation of human motor activity.

The modern concept of the mechanism of conditional

connection closure (Hazy et al., 2009) assumes that

the association of excitation foci corresponding to the

conditioned and unconditioned stimuli can occur both

at the level of the cortex and at the level of the

subcortex. With continued flowing along specific

paths to a certain limited cortical focus of afferent

impulses, gradually generalized excitation is

concentrated in this focus. Then it gives a significant

part of its influence on the construction of movement

to the underlying foci of excitation, which have the

advantage that afferent proprioceptive impulses

continue to flow to them.

Figure 4: The distribution of time delays between the

moment of presentation of the sound signal and the detected

time moment (precursor) of movement beginning. The

distribution is based on the results of one experimental

session consisting of a series of repetitions of the

movement. The results were obtained for the optimal

threshold value R for the derivative of the smoothed EMG

signal. The blue curve is Gaussian kernel density estimate.

Recent studies indicate the presence of delays in the

activation of sensorimotor processing in the human

brain associated with the phases of formation,

recognition of the stimulus, categorization of

response, decision making and reaction of afferent

neurons that have times comparable to those obtained

in this work, although they take less values due to the

specifics of the experiment (Melnik et al., 2017;

Asakawa et al., 2014).

It should be noted, that the distribution in Fig. 4 is

rather well approximated by the Gaussian-like

distribution. The time it takes for the pulse to travel

from the brain to the muscle and the reaction time of

the muscle are approximately constant for all

repetitions of movement. Consequently, the noise

component, which determines the form of distribution

of the time delay h (Fig. 4), is a consequence of the

processes occurring in the brain, when processing the

The Approach to the Detection of the Movement Precursor by Electromyographic Signals

279

stimulus and the generating the control signal. Indeed,

the effect of “brain noise” was discovered in

(Pisarchik et al., 2019; Runnova et al., 2016). Thus, it

can be assumed that the initial state of the brain before

each act of movement is different and is determined

by the processes in the brain at this moment, which

determines the noise nature of the distribution.

4 CONCLUSIONS

In the work, a method was proposed that allows to

automatically determine the precursor of the

movement beginning, based on the analysis of EMG

signals. It was found out that in the case when the

motion begins on the sound signal, the moment of the

start of motion is detected some time after the signal,

the distribution of which is approximated fairly well

by the Gaussian-like distribution. Possible causes and

background of the obtained results are discussed. The

obtained results can be used to isolate the phases of

“movement planning” and contribute to solving a

number of applied problems associated with

improving the quality of life of people and with

development of human-machine interfaces. The

proposed technique has the potential for application

in human-machine interfaces.

ACKNOWLEDGEMENTS

This work has been supported by Russian Science

Foundation (Grant 17-72-30003).

REFERENCES

Wood, G., Kober, S., Witte, M., Neuper, C., 2014. On the

need to better specify the concept of “control” in

brain-computer-interfaces/neurofeedback research.

Front Syst Neurosci. Vol. 8, pp. 171.

https://doi.org/10.3389/fnsys.2014.00171.

Hayashibe, M., Guiraud, D., Pons, J., Farina, D., 2015

Editorial: biosignal processing and computational

methods to enhance sensory motor neuroprosthetics.

Front Syst Neurosci. Vol. 9, pp. 434.

Maksimenko V.A., Pavlov A., Runnova A.E., Nedaivozov

V., Grubov V., Koronovskii A., Pchelintseva S.V.,

Pitsik E., Pisarchik A.N., 2018. Hramov A.E. Nonlinear

analysis of brain activity, associated with motor action

and motor imaginary in untrained subjects. Nonlinear

Dynamics. Vol. 91(4) pp. 2803-2817 DOI:

10.1007/s11071-018-4047-y.

Mondini, V., Mangia, A., Cappella, A., 2018. Single-

session tDCS over the dominant hemisphere affects

contralateral spectral EEG power, but does not enhance

neurofeedback-guided event-related desynchronization

of the non-dominant hemisphere's sensorimotor

rhythm, PLoS One. Vol. 13(3), e0193004.

Reis, P., Hebenstreit, F., Gabsteiger, F., von Tscharner, V.,

Lochmann, M., 2014. Methodological aspects of EEG

and body dynamics measurements during motion.

Frontiers in Human Neuroscience. Vol. 8, pp. 156.

Rouillard, J., Duprèsa, A., Cabestainga, F., Leclercqb, S.,

Bekaerta, M., Piaua, C., Vannobela, J., Lecocq, C.,

2015. Hybrid BCI coupling EEG and EMG for severe

motor disabilities, Procedia Manufacturing. Vol. 3, pp.

29–36.

Basmajian, J., 1979. Muscle alive, their functions are

revealed by electromyography. Williams and Wilkins,

4th edition.

De Luca, C., 2010. Filtering the surface EMG signal:

Movement artifact and baseline noise contamination,

Journal of Biomechanics. Vol. 43, pp. 1573–1579.

Kastalskiy, I., Mironov, V., Lobov, S., Krilova, N.,

Pimashkin, A., Kazantsev, V., 2018. Neuromuscular

Interface for Robotic Devices Control. Computational

and Mathematical Methods in Medicine. 8948145.

Hazy, T., Frank, M., O’Reilly, R., 2009. Neural

Mechanisms Supporting Acquired Phasic Dopamine

Responses in Learning: An Integrative Synthesis.

Neuroscience and Biobehavioral Reviews. Vol. 34(5),

pp. 701–720.

Melnik, A., Hairston, W., Ferris, D., König, P., 2017. EEG

correlates of sensorimotor processing: independent

components involved in sensory and motor processing.

Scientific Reports. Vol. 7, pp. 4461.

Asakawa, T., Muramatsu, A., Hayashi, T., Urata, T., Taya

M., Mizuno-Matsumoto Y., 2014. Comparison of EEG

propagation speeds under emotional stimuli on

smartphone between the different anxiety states.

Frontiers in Human Neuroscience. vol. 8, pp. 1006.

Pisarchik, A.N., Chholak, P., Hramov, A.E., 2019. Brain

noise estimation from MEG response to flickering

visual stimulation. Chaos, Solitons & Fractals: X.

Vol. 1, pp. 100005. DOI: 10.1016/j.csfx.2019.100005

Runnova, A.E., Hramov, A.E., Grubov, V.V., Koronovskii,

A.A., Kurovskaya, M.K., Pisarchik, A.N., 2016.

Theoretical background and experimental

measurements of human brain noise intensity in

perception of ambiguous images. Chaos, Solitons and

Fractals. Vol. 93, pp. 201-206.

ICINCO 2019 - 16th International Conference on Informatics in Control, Automation and Robotics

280