Me hods o Powe Line In e e ence Elimina ion in EMG Signals
Ma ina Lad o a1,a, Radek Ma inek1,b, Jan Nedoma2,c*, Ma cel Fajkus2,d
1Depa men o Cybe ne ics and Biomedical Enginee ing, Facul y o Elec ical Enginee ing and
Compu e Science, VSB–Technical Uni e si y o Os a a, Os a a, Czech Republic
2Depa men o Telecommunica ions, Facul y o Elec ical Enginee ing and Compu e Science
VSB–Technical Uni e si y o Os a a, Os a a, Czech Republic
ama ina.lad o a@ sb.cz, b[email p o ec ed], c*jan.nedoma@ sb.cz, d[email p o ec ed]
Keywo ds: Elec omyog aphy, Biological signal p ocessing, Powe Line, No ch il e , Adap i e
Noise Cancelle , Wa ele T ans o m.
Abs ac . Elec omyog am (EMG) eco dings a e o en co up ed by he wide ange o a i ac s,
which one o hem is powe line in e e ence (PLI). The s udy ocuses on some o he well-known
signal p ocessing app oaches used o elimina e o a enua e PLI om EMG signal. The esul s a e
compa ed using signal- o-noise a io (SNR), co ela ion coe icien s and Bland-Al man analysis o
each es ed me hod: no ch il e , adap i e noise cancelle (ANC) and wa ele ans o m (WT). Thus,
he powe o he emaining noise and shape o he ou pu signal a e analysed. The esul s show ha
he ANC me hod gi es he bes ou pu SNR and lowes shape dis o ion compa ed o he o he
me hods.
1. In oduc ion
Elec omyog am (EMG) is a ype o diagnos ic me hod, which allows measu ing o he biological
signal ha o igina es due o he skele al muscle ac i i y. The echnique o en consis s o he
s imula ion o he ele an mo o ic and pe iphe al ne es and simul aneous measu ing o he EMG
signal. The measu ing can be ei he in asi e o non-in asi e, moni o ing he ac i i y o single muscle
ibe , one mo o uni o whole muscle. The signal gi es us he in o ma ion which can be used o he
diagnose o muscula and neu omuscula disease, walk analysis, ehabili a ion o con ol he limb
p os hesis.
The EMG signal seems o be e y complica ed in i s analysis, because o i s non-s a iona i y, which
complica es he use o he il a ion in a equency domain, and wide equency ange om 10 Hz up
o 1 kHz. The EMG signal, such as e e y biological signal, is in luenced by many ypes o in e e ence
and a i ac s. The powe line in e e ence (PLI) on equency 50 Hz (in Eu ope) o 60 Hz (in No h
Ame ica), which is objec o his pape , b ings he mo e signi ican p oblem in EMG signal han in
he case o he o he biological signals due o he signal equency ange – he moni o ed ange o
o he biological signals (e.g. ECG, EEG) could be educed unde 50 Hz, which enables he use o he
low pass il e . In he case o EMG, i is no possible o use he low pass il e because o i s equency
up o 1 kHz and i is necessa y o use he no ch il e o some o he o he ypes o il a ion, e.g.
wa ele ans o m (WT) o adap i e il e ing [1,2,3].
The pape compa es he esul s o he well-known me hods o EMG signal il a ion (no ch il e ,
adap i e noise cancelle (ANC) and WT) be ween he o iginal a i icially gene a ed EMG signal and
he il e ed signal a e mixing wi h he 50 Hz PLI. The MATLAB so wa e was used o
implemen a ion o he il a ion me hods, and he co ela ion and signal- o-noise a io (SNR) we e
compu ed o he e ec i eness e alua ion o each me hod.
2. Ma e ial and Me hods
The so wa e o biomedical signal p ocessing was implemen ed in MATLAB. I ep esen s he
ool o es ing and compa ing he well-known p ocessing and il a ion me hods used in one-
dimensional biological signal p ocessing. The so wa e allows loading signals om .ma ile, adding
he basic ypes o in e e ence (high and low- equency in e e ence, powe line in e e ence, whi e
Jou nal o Biomime ics, Bioma e ials and Biomedical Enginee ing Submi ed: 2018-10-19
ISSN: 2296-9845, Vol. 40, pp 64-70 Re ised: 2018-10-26
doi:10.4028/www.scien i ic.ne /JBBBE.40.64 Accep ed: 2018-10-26
© 2019 T ans Tech Publica ions, Swi ze land Online: 2019-02-22
This is an open access a icle unde he CC-BY 4.0 license (h ps://c ea i ecommons.o g/licenses/by/4.0/)
noise o ends) o he signal and signal p ocessing including FIR and IIR il e ing, ANC, de i a ion,
WT, independen and p incipal componen analysis, e c. A e each change o he p ocessed signal
(e.g. adding in e e ence o il e ing), he SNR and co ela ion coe icien s a e calcula ed o compa e
he e ec i eness o each signal p ocessing me hod. The signals a e e alua ed also by he g aphical
way – ime wa e o ms, equency spec a and spec og ams a e shown o each ype o signal –
o iginal, co up ed and il e ed (Fig. 1).
The pape a ge s on he EMG signal p ocessing. The h ee me hods o PLI emo al a e discussed in
de ail in he ollowing subsec ions.
2.1 No ch il e
The no ch il e is he simples me hod o emo ing he ixed na ow bandwid h in e e ence om
he signal. The ma hema ical exp ession o i s ans e unc ion is de ined in (1).
𝐻𝐻(𝑧𝑧)=𝑏𝑏0
1−2𝑐𝑐𝑐𝑐𝑐𝑐
ω
0𝑧𝑧−1+𝑧𝑧−2
1−2𝑟𝑟𝑐𝑐𝑐𝑐𝑐𝑐
ω
0𝑧𝑧−1+𝑟𝑟2𝑧𝑧−2, (1)
whe e ω0 is he cen al equency o he in e e ence and he wid h o he no ch a -3 dB is 2(1- )
ad/s.
Because he PLI equency and i s ha monics a e loca ed wi hin he bandwid h o EMG, no ch il e
emo es PLI as well as he componen s o EMG signal loca ed a his equency and ew equencies
a ound. The e o e, he il e causes he in o ma ion loss. Fo he dis o ion minimiza ion, na ow band
o he il e should be used [1,3].
Fig. 1. MATLAB so wa e GUI. So wa e enables co up ing o he signal wi h many ypes o he
in e e ences and a e ac s, i s il e ing and o he p ocessing. So wa e was implemen ed as he ool
o es ing me hods o biological signal p ocessing and hei compa ison.
2.2 Adap i e noise cancelle
The me hod o ANC il a ion exploi s wo inpu signals [7,8,9]. The p ima y inpu is he
combina ion o he o iginal signal and he noise, whose emo al is he a ge o he me hod. The
seconda y inpu is he syn he ic e e ence, which is in co ela ion wi h he noise. The il e modi ies
he e e ence so ha he signal e y simila o he noise is gained and he ou pu signal is ob ained as
i s di e ence wi h he inpu signal.
Jou nal o Biomime ics, Bioma e ials and Biomedical Enginee ing Vol. 40 65
The ou pu is connec ed o he locked loop o he il e so ha minimal ou pu powe and also a
small powe o he noise in ou pu signal a e ob ained. Fo he as op imiza ion o he inpu signal
weigh s, he RLS (Recu si e Leas Squa e) il e [14,15,16] which uses he sum o he squa es o he
e o signal (de ia ion o he e e ence signal and he inpu signal) as a c i ical unc ion, is used [1,3].
2.3 Wa ele ans o m
Because o many limi a ions o he Fou ie ans o m used in he p ocessing o he non-s a iona y
signals, he me hods o he ime- equency p ocessing a e ound. The idea o he WT is he sui able
changing o he ma e nal wa ele ’s wid h in ime and o gain he op imal a io be ween ime and
equency dis inguishabili y [10,11,12].
The WT echnique is e y o en used in he p ocessing o he non-s a iona i y signals bu also in
o he cases, because o he ex ensi e selec ion o he ma e nal wa es, whose shape o en esponds o
hei main cha ac e s (e.g. QRS complex o ECG) [4,5]. Acco ding o [6], many ypes o ma e nal
wa ele s we e es ed and h ee o hem we e e alua ed in he s udy (Db6, Sym4 and Meye wa ele ).
3. C i e ia o Pe o mances
The h ee me hods discussed in he p e ious sec ion a e applied o he co up ed EMG signal. The
c i e ia used o he pe o mance analysis o each me hod a e SNR and shape dis o ion o he ou pu
signal.
3.1 Signal gene a ion
The simula ed EMG signal was gene a ed using LabView Biomedical oolki [18] and has he
ollowing p ope ies:
• maximum ampli ude 4 mV,
• con ac ion du a ion 1 s,
• equency bandwid h 20 – 300 Hz,
• sampling equency 1 kHz.
Fi s , he 50 Hz PLI wi h ampli ude 0.4 mV is added o he gene a ed EMG signal. This signal has
he SNR alue o 6.459 dB. Then, he co up ed EMG signal is il e ed using he me hods discussed
abo e, and SNR and co ela ion coe icien wi h he o iginal signal a e calcula ed. Fo he g aphic
ep esen a ion, also he Bland-Al man analysis is shown o each me hod.
3.2 Signal- o-noise a io
The alue o SNR is de ined as he a io o he simula ed signal powe (PS) o he powe o he
emaining in e e ence (PN) acco ding o (2).
𝑆𝑆𝑆𝑆𝑆𝑆 =10 ∙𝑙𝑙𝑙𝑙𝑙𝑙
𝑃𝑃
𝑠𝑠
𝑃𝑃𝑁𝑁
. (
2)
The SNR alue is calcula ed o he signal a e adding he in e e ence and hen, o each
il e ing me hod o compa e, how much he me hod was success ul. Wi h he inc easing SNR, he
powe o he noise is smalle , and he il e ing p ocess is mo e quali y.
66 Jou nal o Biomime ics, Bioma e ials and Biomedical Enginee ing Vol. 40
3.3 Co ela ion coe icien
The co ela ion coe icien is also compu ed o he signal wi h in e e ence and he pai o he ou pu
signal and he o iginal simula ed signal. The alue o he co ela ion coe icien anged om 0 o 1
shows he shape simila i y o he il e ed signal and he o iginal signal. Highe co ela ion coe icien
means a lowe signal’s shape dis o ion.
3.4 Bland-Al man analysis
Bland-Al man g aph [12,13] is used mainly o compa e he esul s o all me hods. The cha shows
he ela ion be ween he esul o he signal il a ion and he o iginal signal. Bland-Al man analysis
compa es wo a iables in hei di e ence and a e -age, which a e compu ed o each sample. The
esul ing g aph is he dependence o his di e ence and a e age. The a e age alue (AVG) and he
s anda d de ia ion (SD) a e compu ed and shown in he g aph as h ee ho izon al lines (AVG, AVG
± 1.96·SD).
Mo e samples loca ed a ound he AVG ho izon al line shows he mo e accu a e ela ion be ween he
o iginal and he ou pu signal.
4. Resul s
Each o he h ee me hods men ioned in sec ion 2 gained di e en esul s ha a e p esen ed in his
sec ion.
No ch il e . The classical IIR no ch il e [17] elimina es he spec al lines o he selec ed
equency componen s bu also educes he powe o he nea by equencies. The su icien
a enua ion o he PLI equency can be ob ained by inc easing he wid h o he s opband. Howe e ,
he wide s opband causes mo e in o ma ion emo al and he signal is dis o ed. In his s udy, he
il e o de s 1, 5 and 10 we e used. In he case o he o de 1, SNR descended beside he o iginal
alue bu using he highe o de SNR inc eased. On he o he hand, he co ela ion achie ed qui e
high alue in all cases.
Adap i e noise cancelle . In he case o he ANC il a ion me hod, SNR in-c eased e y highly,
and he co ela ion was p ese ed. Because ANC wo ks wi h he e e ence which in his case en i ely
copies he in e e ence signal, he signal is no dis o ed in any way and only he PLI is emo ed. The
o de o he adap i e il e was se o 1, 5 and 10. The highes SNR was gained in he case o he o de
1, wi h inc easing o de SNR dec eased.
Wa ele ans o m. F om he a ailable ypes o ma e nal wa es in MATLAB so wa e, he Db6,
Sym4 and Meye wa e gained be e esul . Howe e , SNR dec eased o ze o and co ela ion
coe icien eached only abou 0.24. Though WT is a ecommended me hod o he p ocessing o he
non-s a iona y signals, in his case his me hod eached he lowes esul s in SNR and co ela ion
because o he elimina ion o he highe equencies o he signal and he in o ma ion loss.
Jou nal o Biomime ics, Bioma e ials and Biomedical Enginee ing Vol. 40 67
Fig. 2. Time wa e o ms and equency spec a o he signal be o e (a, b) and a e (c, d) il e ing
using ANC me hod.
Table 1. Pe o mance esul s o each me hod.
Me hod SNR [dB]
Co ela ion
coe icien
No ch
1
5.847
0.8601
5
12.11
0.9687
10
15.33
0.9852
ANC
1
38.25
0.9999
5
35.78
0.9999
10
36.36
0.9999
WT
Db6
0.2617
0.2418
Sym4
0.2457
0.2345
Meye
0.2437
0.2336
68 Jou nal o Biomime ics, Bioma e ials and Biomedical Enginee ing Vol. 40

Fig. 3. Bland-Al man g aphic analysis o each me hod shows i s acco dance wi h he o iginal EMG
signal. The g aph o he ANC me hod indica es he bes esul because o he lowes di e ence alues
in all samples o he signal.
The esul s indica e, ha om he selec ion o me hods used in his s udy, he ANC il a ion me hod
is applicable in acco dance o elimina e PLI om EMG signal. I gained he bes esul s and he
co ela ion is s ill close o 1, so i has he minimum e ec on he shape o he signal.
5. Conclusion
In he s udy, h ee me hods used o educing PLI in he EMG signal a e compa ed. Fo he
analysis o hei pe o mance, SNR and co ela ion coe icien s a e compu ed as indexes o each
echnique. The SNR alues o he signal con aining PLI and he ou pu signal show he accu acy o
each me hod. Conce ning SNR and co ela ion coe icien s, ANC il e p o es o be he mos sui able
me hod o EMG signal p ocessing in he case o PLI emo al.
Acknowledgmen
This a icle was suppo ed by he Minis y o Educa ion o he Czech Republic (P ojec No.
SP2018/170). This wo k was suppo ed by he Eu opean Regional De elopmen Fund in he Resea ch
Cen e o Ad anced Mecha onic Sys ems p ojec , p ojec numbe
CZ.02.1.01/0.0/0.0/16_019/0000867 wi hin he Ope a ional P og amme Resea ch, De elopmen and
Educa ion.
Re e ences
[1] R. M. Rangayyan, Biomedical signal analysis, New Je sey: John Wiley & Sons. 33 (2015).
[2] E.N. B uce, Biomedical signal p ocessing and signal modeling, New Yo k: Wiley. (2001) 335–
336.
[3] S.D. Soedi djo, K. Ullah, R. Me le i, Powe line in e e ence a enua ion in mul i-channel sEMG
signals: Algo i hms and analysis, In: EMBC (2015) 3823–3826.
[4] M. Reaz, M.S. Hussain, F. Mohd-Yasin, Techniques o EMG signal analysis: de ec ion,
p ocessing, classi ica ion and applica ions, Biological P ocedu es Online. 8 (2006) 11–35.
[5] P.S. Addison, The illus a ed wa ele ans o m handbook: in oduc o y heo y and applica- ions
in science, enginee ing, medicine and inance, CRC p ess (2017).
Jou nal o Biomime ics, Bioma e ials and Biomedical Enginee ing Vol. 40 69
[6] M. Hussain, M. Reaz, F. Mohd‐Yasin, M. Ib ahimy, Elec omyog aphy signal analysis us-ing
wa ele ans o m and highe o de s a is ics o de e mine muscle con ac ion, Expe Sys ems. 26
(2009) 35–48.
[7] R. Ma inek, J. Konecny, P. Koudelka, J. Zidek, H. Naze an, Adap i e op imiza ion o con- ol
pa ame e s o eed- o wa d so wa e de ined equaliza ion, Wi eless Pe sonal Commu-nica ions. 95
(2017) 4001–4011.
[8] R. Ma inek, M. Kelna , J. Vanus, P. Koudelka, P. Bilik, J. Kozio ek, J. Zidek, Adap i e noise
supp ession in oice communica ion using a neu o- uzzy in e ence sys em, In: Tele-communica ions
and Signal P ocessing. (2015) 382–386.
[9] A. Sugiyama, Adap i e noise cancelle wi h wo SNR es ima es o s epsize con ol, In: Consume
Elec onics. (2018) 1–2.
[10] R. Ma inek, R. Kahanko a, J. Nedoma, M. Fajkus, K. Chole o a, Fe al ECG P ep o-cessing
Using Wa ele T ans o m, In: P oceedings o he 10 h In e na ional Con e ence on Compu e
Modeling and Simula ion. (2018) 39–43.
[11] P.S. Addison, The illus a ed wa ele ans o m handbook: in oduc o y heo y and applica- ions
in science, enginee ing, medicine and inance, CRC p ess (2017).
[12] R. Ma inek, R. Kahanko a, J. Jezewski, R. Ja os, J. Mohylo a, M. Fajkus, M., J. Nedoma, H.
Naze an, Compa a i e E ec i eness o ICA and PCA in Ex ac ion o Fe al ECG F om Abdominal
Signals: Towa d Non-in asi e Fe al Moni o ing. F on ie s in physiology. 9 (2018).
[13] D. Gia a ina, Unde s anding bland al man analysis. Biochemia medica: Biochemia medica.
(2015) 141–151.
[14] R. Ma inek, J. Zidek, P. Bilik, J. Manas, J. Kozio ek, H. Wen, The use o lms and ls adap i e
algo i hms o an adap i e con ol me hod o ac i e powe il e , Ene gy and Powe Enginee ing. 5
(2013) 1126–1133.
[15] R. Kahanko a, R. Ma inek, P. Bilik, Non-in asi e Fe al ECG Ex ac ion om Ma e nal
Abdominal ECG Using LMS and RLS Adap i e Algo i hms, In: In e na ional A o-Eu opean
Con e ence o Indus ial Ad ancemen . (2016) 258–271.
[16] R. Kahanko a, R. Ma inek, P. Bilik, Fe al ECG ex ac ion om ab-dominal ECG using RLS
based adap i e algo i hms, In: Ca pa hian Con ol Con e ence. (2017) 337–342.
[17] H.K. Jayan , K. Rana, V. Kuma , S. Nai , P. Mish a, E icien IIR no ch il e design using
Minimax op imiza ion o 50Hz noise supp ession in ECG, In: Signal P ocessing, Compu ing and
Con ol. (2015) 290–295.
[18] A. Deshmukh, Y. Gandole, ECG ea u e ex ac ion using NI LAB-VIEW biomedical wo kbench,
In e na ional Jou nal o Recen Scien i ic Resea ch, 6 (2015) 5603–5607.
70 Jou nal o Biomime ics, Bioma e ials and Biomedical Enginee ing Vol. 40