Clinical StudyComputation and Evaluation of Features of SurfaceElectromyogram to Identify the Force of Muscle Contractionand Muscle Fatigue
Sridhar P Arjunan1 Dinesh K Kumar1 and Ganesh Naik2
1 Biosignals Lab School of Electrical and Computer Engineering RMIT University GPO Box 2476 Melbourne VIC 3001 Australia2 Faculty of Engineering and Information Technology (FEIT) University of Technology Sydney (UTS) Sydney NSW 2007 Australia
Correspondence should be addressed to Sridhar P Arjunan sridhararjunanrmiteduau
Received 4 March 2014 Accepted 20 May 2014 Published 4 June 2014
Academic Editor Terry K Smith
Copyright copy 2014 Sridhar P Arjunan et al This is an open access article distributed under the Creative Commons AttributionLicense which permits unrestricted use distribution and reproduction in any medium provided the original work is properlycited
The relationship between force of muscle contraction and muscle fatigue with six different features of surface electromyogram(sEMG) was determined by conducting experiments on thirty-five volunteers The participants performed isometric contractionsat 50 75 and 100 of their maximum voluntary contraction (MVC) Six features were considered in this study normalisedspectral index (NSM5) median frequency root mean square waveform length normalised root mean square (NRMS) and increasein synchronization (IIS) index Analysis of variance (ANOVA) and linear regression analysis were performed to determine thesignificance of the feature with respect to the three factors muscle force muscle fatigue and subjectThe results show that IIS indexof sEMG had the highest correlation with muscle fatigue and the relationship was statistically significant (119875 lt 001) while NSM5associated best with level of muscle contraction (MVC) (119875 lt 001) Both of these features were not affected by the intersubjectvariations (119875 gt 005)
1 Introduction
Surface electromyogram (sEMG) is the recording of theelectrical activity that is associated withmuscle activation [1]sEMG is the interferential summation of tissue-filteredmotorunit action potentials (MUAP) generated by active motorunits and represents a pattern characterizing the general stateof the muscle examined [2 3]
Strength of muscle contraction is dependent on the num-ber of active motor units their size the rate of stimulation ofthe motor units and the type of muscle fibres The ability ofthe muscle to contract and produce force can diminish oversustained contraction and when it is localized to a muscle orgroup ofmuscles that is referred to as localizedmuscle fatigue[4 5] and is also closely associated with sEMG Numerousstudies [6ndash14] have reported the relationship of sEMG withforce of muscle contraction and localized muscle fatigue
Various features of sEMG such as root mean square(RMS) [6] median frequency [7] wavelet transforms [8 9]
fractal dimension [10 11] normalized spectral moments [1213] and increase in synchronization (IIS) index [14] havebeen related to parameters of muscle contraction such asforce and muscle fatigue However there are a number ofcompounding factors such as force of contraction onset ofmuscle fatigue length of the muscle tissue properties andexternal factors such as noise and intersubject variations thatinfluence sEMG and thus sEMG is considered to be suitablefor onlymeasuring the relative change in themuscle state [15ndash17]
Surface EMG is noninvasive and easy to record signaland machine based estimation of force of muscle contractionor for assessing muscle fatigue it will have large number ofrehabilitation and other applications However while a num-ber of studies have identified different features of sEMG anddemonstrated the association of these with force and fatigueno study has compared the relationship of these features andevaluated these features for automated estimation of force andfatigue from sEMG
Hindawi Publishing CorporationBioMed Research InternationalVolume 2014 Article ID 197960 6 pageshttpdxdoiorg1011552014197960
2 BioMed Research International
The aim of this study was to experimentally determinethe most suitable of the currently used sEMG features thatcan be implemented for machine based sEMG analysis toestimate muscle force and fatigue This research has exper-imentally studied six well-accepted features of sEMG andanalyzed the relationship of each of these with force ofmusclecontraction and with muscle fatigue Linear regression andanalysis of variance (ANOVA) were performed to comparethe relationship of each of these features with the force ofmuscle contraction and muscle fatigue The significance ofthis study is that it has shown a comparison between thevarious features of sEMG that have been reported in literatureand has identified the relationship of differences due to threefactors subject force and fatigue
2 Materials and Methods
21 sEMG Recording sEMG signals were recorded using theproprietary Delsys (Boston MA USA) sEMG acquisitionsystem The system supports bipolar recording and has gainof 1000 CMRR of 92 dB and bandwidth of 20ndash450Hz with12 dBoctave roll-off The sampling rate was fixed at 1000samplessecond and the resolutionwas 16 bitssample Activebipolar electrodes (Delsys Boston MA USA) having twosilver bars (1mm wide and 10mm long) mounted directly onthe preamplifier with fixed interelectrode distance of 10mmwere used for recording sEMG
The experiments were conducted on the biceps brachiimuscle because this muscle is superficial and is the primemover for the ankle joint and the muscle fibers run parallelto the surface Two bipolar electrodes were placed above themotor points of the short head of the biceps brachii muscleand inline between the anticubital fossa (depression in thefront of the elbowmdashlateral to the biceps brachii tendon) andthe acromion process (part of the scapula which extends overthe shoulder) at 13rd distance from the anticubital fossaThedistance between the two bipolar electrodes was maintainedat 2 cm Reference electrode was placed on the dorsal sectionand under the elbow Prior to electrode placement the skinarea was cleaned with alcohol swabs and lightly exfoliatedwith paper towel to reduce skin impedance and ensure goodadhesion of the electrodes
22 Experimental Protocol Thirty-five healthy subjects (22male and 13 female aged 22ndash35 years) consented to par-ticipate in these trials The participants with the followingcriteria were excluded (1) arthritis (eg osteoarthritis andrheumatoid arthritis) (2) neuromuscular disorders includingcollagen disorders and nonarticular rheumatism includingfibro myalgia or seizure disorders (3) any recent injuries ofthe hand wrist or arm
The experiments were approved by theUniversityHumanEthics Committee The experiments were performed inaccordance with Declaration of Helsinki of 1975 as revisedin 2004 During the experiments the volunteers were seatedsuch that their feet were flat on the floor the upper arm restedhorizontally on an adjustable desk and the forearm wasvertical (refer to Figure 1) with the elbow at 90 degrees Awall
Force sensor
Force measurement EMG recording unit
Force EMG display
Figure 1 Illustration of the experimental setup
mounted force sensor (S-type force sensormdashINTERFACESM25) was attached to a comfortable hand sized ring with aflexible steel wire padded for comfort and the ring was heldon the palm of the participantThe output of the force sensorwas recorded alongside the sEMG recording on the Delsysacquisition system Before the start of the experiment theexperimental procedures were explained to the participantsand trial runs were performed to familiarize them with theexperiment
Determining Maximum Voluntary Contraction (MVC) Todetermine the maximal voluntary contraction (MVC) threemaximal contractions of 5 seconds were performed with 120seconds rest time between each effortThe participants pulledon the ring and the force of contraction was recorded Theforce output was displayed on the screen and the participantswere encouraged to exert their maximal muscle force andsteadily maintain the force The average of the three readingswas considered to be the MVC If there were any outliers theexperimentwas repeated Preliminary experiments were con-ducted to check the influence of triceps and brachioradialismuscles on biceps during the contraction by recording sEMGof these muscles If the muscle activity of these muscles wasgreater than 5 of MVC of biceps (estimated using RMS ofsEMG) the participants were reseated and assisted in restingthese muscles
After confirming that the triceps and brachii were silentand the activity was concentrated on the biceps muscle theexperiments were conducted during which the participantsperformed three sets of isometric contractions at 50 75and 100 MVC respectively Force of contraction was dis-played on the computer to give them feedback and assistthem in maintaining steady force Participants were asked toperform the contractions as long as they could and until theyexperienced fatigue and pain The trial was terminated whenthe subject was unable to exert the required force and theexerted force dropped below 80 of the target or when thesubjects experienced pain whichever occurred earlier Thepain was subjectively measured using a numeric rating pain
BioMed Research International 3
10
8
6
4
2
0
minus2
minus4
minus6
minus8
0 1 2 3 4 5 6 7 8
Mag
nitu
de (V
)
Time in samples
times10minus5
times104
Figure 2 A sample of the recorded raw sEMG signal (time insamples)
index scale (PIS) [5 18] with range from 0 to 10 and withPIS of 0 corresponding to ldquono painrdquo and 10 corresponding toldquomaximum painrdquo Subjects were informed of the PIS beforethe start of the experiment and were given examples to helpthem understand the scale The subjects were requested andreminded to report the pain only in the biceps muscle Ascore of 8 and above corresponded to the limit of muscleendurance
The total duration of each contraction was referred toas the endurance period This was found to be differentfor different participants and for different levels of musclecontraction To allow a comparison between different exper-iments the time axis was normalized such that the startof the experiment corresponded to T
1and the end of the
exercise was labeled T119864 The participants were given a rest
period of 60 minutes between each contraction but as longas they required The participants were questioned such thatthey reported the muscle to be pain-free and rested Theexperiments performed in this study were similar to ourearlier published protocol [14] A sample raw EMG signalrecorded in the experiment has been shown in Figure 2
23 Data Analysis Data analysis was performed offline onMATLAB 2009a software environment (The MathWorksInc Natick Massachusetts USA) The first step was thetemporal segmentation of the recordings and this was fol-lowed by computation of the six features Regression analysiswas performed to determine the linearity of the relationshipbetween each of the features and force of contraction FinallyN-factor analysis of variance (ANOVA) was computed todetermine the statistical significance of the relationship
231 Temporal Segmentation All sEMG recordings weredivided into 1 second long segments (1000 samples) withoverlap of 100 samples using amovingwindowThe average ofthe feature of all the windows in the corresponding segment
was computed and labeled according to the segment numberThe first segment corresponding to the start of the exercisewas labeled as that of the rested muscle while the finalsegment corresponded to the fatigued muscle at the limit ofendurance
232 Computation of Features The following features werecomputed
(i) Normalized SpectralMoments (NSM5)The spectral fatigueindices (FInsm5) proposed by Dimitrov et al [12] and Dim-itrova et al [13] are the measure of the normalised spectralmoments as follows
FInsm5 =int1198912
1198911
119891minus1sdot PS (119891) sdot 119889119891
int1198912
1198911
1198915 sdot PS (119891) sdot 119889119891 (1)
where PS(f) is the EMG power spectrum 1198911= 8Hz 119891
2=
500Hz and ldquosdotrdquo represents the multiplication factor (see [1213] for further details)
(ii) Median Frequency (MDF) Median frequency is theparticular frequency that divides the power spectrum intotwo sections of equal areas (see [19 20] for details)
(iii) Root Mean Square (RMS) Root mean square (RMS) isthe quadraticmean and a statisticalmeasure of themagnitudeof a time varying signal and is computed using the followingequation
RMS = radic 1119873
119873
sum
119894=1
1199092
119894 (2)
where119873 is the number of samples in the segment and 119909 is thesEMG signal
(iv) Normalised RMS (NRMS) RMS of sEMG for eachparticipant (at 75 and 50 MVC) was normalized withrespect to the RMS of sEMG corresponding to their MVCof the rested muscle and was computed using the followingequation
NRMS =RMS75 50MVC
RMSMVC (3)
(v) Waveform Length (WL) Wavelength is the measure ofthe length of the signal and is computed using the followingequation
WL = 1119873
119873minus1
sum
119894=1
1003816100381610038161003816119909119894+1 minus 1199091198941003816100381610038161003816
(4)
where119873 is the number of samples in the segment and 119909 is thesEMG signal (in samples)
(vi) Increase in Synchronization (IIS) Index Increase in syn-chronization (IIS) index is the measure of independencebetween two signals In this study IIS index was computed
4 BioMed Research International
using the EMG recordings from the two channels (twosensors) [14] using 1 second window length similar to otherfeatures The computation of IIS index has been explainedin detail in our earlier publications [14 21ndash23] The signalwas filtered into four narrow subband components of equalbandwidth (band pass filter with 125Hz frequency band)Independent component analysis (ICA) was performed oneach subband component and the resultant 119899 = 4 unmixingor separating square matrices119882
119899 The global matrix 119866 was
estimated as the product of the 119899th unmixing matrix and theinverse of the (119899 + 1)th unmixing matrix Average of 119866of all the time windows in the segment was computed toobtain 119866 and IIS index corresponding to the segment wasobtained by computing log 119866 the normalized determinantof the global matrix
233 Regression Analysis To determine the relationshipbetween the level of contraction (force) and the featuresregression analysis was performed For this analysis thefeatures computed from the sEMG were recorded during theinitial state of the muscle contraction Preliminary analysisshowed that the relationships were linear in nature Basedon this criterion the relationship between the level of con-traction during initial state of maintaining the force and thefeatures was computed using linear regression analysis with95 confidence intervals
234 Statistical Analysis The statistical significance of theeffect and the relationship between the different factors oneach of the six features of sEMG was studied Three-wayanalysis of variance (ANOVA) with interaction conductedfor each of the 6 features with 95 confidence interval (119875 lt005) was performed Kurtosis measures and skew tests wereperformed to check and confirm the underlying assumptionsof ANOVA in analysing the data The three factors were(i) difference between subjects (ii) level of contraction (MVC) and (iii) muscle fatigue In this study ldquolocalisedmuscle fatiguerdquo was defined as the conditionwhen the subjectwas unable to perform the level of contraction and reporteda high level of discomfort (gt8 on PIS)
The factor subject was considered as a random factorANOVA model was designed to identify the significance ofthe different factors on each of the six features of sEMG Theresult of this analysis would indicate the features that are bestable to identify the effect due to the factor and also determinethe strength of the relationship between the different factors
3 Results
Table 1 is the summary of the results from the three factormain effects using the ANOVA model repeated for each ofthe six features The three factors considered were (i) subject(ii) level of contraction ( MVC) and (iii) localised musclefatigue The effect of localised muscle fatigue was definedas the difference between near the start and near the endof the exercise From Table 1 it is observed that the level ofcontraction ( MVC) as a main effect had significant effectonNSM5 (119875 lt 001) RMS (119875 lt 005) NRMS (119875 lt 005) and
Table 1 Three-way ANOVA for each of the six features with95 confidence interval The three factors and their consideredinteractions are subject level of contraction ( MVC) and musclefatigue factor (initial and final segments of the exercise) forisometric contraction and for the six features of sEMG
FeaturesEffect ofsubject119865-value (119875)
Effect of level ofcontraction ( MVC)119865-value (119875)
IIS index (119875 lt 005) From this table it is also observed thatthe statistical significant relationship of muscle fatigue wasonly with IIS index of sEMG (119875 lt 001) The effect on NMRSat 100 MVC was ignored for the statistical and regressionanalysis because this was the basis for normalisation of therest of the data
Table 2 shows the results of the regression analysis per-formed using the features of sEMG computed from the initialstate of the muscle contraction From this table it is observedthat the relationship of NSM5 of sEMG with muscle forcewas the most linear in comparison with other features (1198772 =096) NRMS also showed a good linearity relationship with1198772 value of 093 The bar plots showing the mean values
(and standard deviation) of the different features with level ofcontraction ( of MVC) are shown in Figure 3 These plotsconfirm the observations from Tables 1 and 2 that NSM5 hasa strong linear relationship with force of muscle contraction
4 Discussion and Conclusion
The association between the sEMG and force of musclecontraction is well accepted Different features of sEMG suchas RMS median frequency normalized spectral momentsand wavelength [7ndash9 11ndash13] have been studied in relationwith the muscle force However none of these studies havereported a comparison between the different features todetermine the statistical significance of these relationshipsnor have these analyzed the linearity of these relationships
BioMed Research International 5
0
50
100
150M
DF
50 75 100Level of contraction (MVC)
(a)
50 75 1000
1
2
3
4
RMS
Level of contraction (MVC)
(b)
0123456
NSM
50 75 100Level of contraction (MVC)
times104
(c)
50 75 100
0
02
IIS
Level of contraction (MVC)
minus06
minus04
minus02
(d)
50 75 1000
05
1
15
Level of contraction (MVC)
NRM
S
(e)
0001002003004
WL
50 75 100Level of contraction (MVC)
minus002
minus001
(f)
Figure 3 Average (plusmn SD) of six features of sEMG and level of contraction for the rested muscle (a) MDF (b) RMS (c) normalised spectralindex (NSM5) (d) IIS index (e) normalised RMS and (f) WL
Such an analysis is important for direct interpretation ofsEMG to estimate the force of muscle contraction
This study has experimentally studied and compared therelationship of muscle force and muscle fatigue with each ofthe six well-accepted features of sEMG Analysis of variance(ANOVA) was conducted to determine the effect of threemain different factors level of muscle contraction ( MVC)muscle fatigue and intersubject variation ANOVA resultsshow that RMS (119875 lt 005) and NSM5 (119875 lt 001) of sEMGwere significantly affected by the level of muscle contraction( MVC) The results also indicate that IIS index (119875 lt 001)was significant in identifying the muscle endurance limit
It is observed from the results that the intersubjectvariation was significant for RMS NRMS andWL while thevariation was not significant for NSM5 and IIS index (119875 gt005) Features such as RMS NRMS and WL are associatedwith the amplitude of the signal and the results indicate thatthere is a significant variation in the amplitude of sEMGbetween subjects However other features that are based onspectrum such as NSM5 or entropy dependent such as IISindex do not have significant differences between subjects
The linearity analysis indicates that the most linearrelationship of force of contraction (ranging from 50 to 100
MVC) was with NSM5 (1198772 = 096) followed by normalizedRMS (1198772 = 093) The linearity relationship between RMSof the signal and force of contraction was poor (1198772 = 071)This was also observed from the plots (Figure 2) The resultssuggest that the relationship between NSM5 a normalizedmeasure of the spectrumof the signal [12] and force ofmusclecontraction ( MVC) is the most significant and linear Theresults also indicate that for biceps NSM5 do not requireany normalization or calibrationThis indicates that NSM5 issuitable to estimate the level of muscle contraction comparedwith other features [6 24] even though this has not beenreported in literature
While earlier studies have identified NSM5 to be a fatigueindex [12 13] this study has shown that NSM5 of sEMGis the measure of force of muscle contraction This maybe attributed to the spectrum of sEMG being significantlyinfluenced by the rate of muscle stimulation and thus withforce of muscle contraction This study has also confirmedthat IIS index [14] is the most suitable indicator of musclebeing at the limit of endurance and is fatiguedThe study alsofound that the advantage of bothNSM5 and IIS was that thesedid not require any normalization
6 BioMed Research International
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
References
[1] J V Basmajian and C J de LucaMuscles Alive Their FunctionsRevealed by Electromyography Williams and Wilkins Balti-more Md USA 1985
[2] R Merletti and S Roy ldquoMyoelectric and mechanical manifes-tations of muscle fatigue in voluntary contractionsrdquo Journal ofOrthopaedic and Sports PhysicalTherapy vol 24 no 6 pp 342ndash353 1996
[3] T J Dartnall M A Nordstrom and J G Semmler ldquoMotor unitsynchronization is increased in biceps brachii after exercise-induced damage to elbow flexormusclesrdquo Journal of Neurophys-iology vol 99 no 2 pp 1008ndash1019 2008
[4] F E Marino M Gard and E J Drinkwater ldquoThe limits toexercise performance and the future of fatigue researchrdquo BritishJournal of Sports Medicine vol 45 no 1 pp 65ndash67 2011
[5] F A BainbridgeThePhysiology ofMuscular Exercise LongmanGreen and Co New York NY USA 3rd edition 1931
[6] S G Boe C L Rice and T J Doherty ldquoEstimating contractionlevel using root mean square amplitude in control subjects andpatients with neuromuscular disordersrdquo Archives of PhysicalMedicine and Rehabilitation vol 89 no 4 pp 711ndash718 2008
[7] RMerletti and P Parker Electromyography JohnWiley amp SonsNew York NY USA 2004
[8] D K Kumar N D Pah and A Bradley ldquoWavelet analysis ofsurface electromyography to determine muscle fatiguerdquo IEEETransactions on Neural Systems and Rehabilitation Engineeringvol 11 no 4 pp 400ndash406 2003
[9] D Moshou I Hostens G Papaioannou and H RamonldquoDynamicmuscle fatigue detection using self-organizingmapsrdquoApplied Soft Computing Journal vol 5 no 4 pp 391ndash398 2005
[10] S P Arjunan and D K Kumar ldquoDecoding subtle forearmflexions using fractal features of surface electromyogram fromsingle and multiple sensorsrdquo Journal of NeuroEngineering andRehabilitation vol 7 no 1 article 53 2010
[11] G Wang X-M Ren L Li and Z-Z Wang ldquoMultifractalanalysis of surface EMG signals for assessing muscle fatigueduring static contractionsrdquo Journal of Zhejiang University AScience vol 8 no 6 pp 910ndash915 2007
[12] G V Dimitrov T I Arabadzhiev K N Mileva J L Bowtell NCrichton andNADimitrova ldquoMuscle fatigue during dynamiccontractions assessed by new spectral indicesrdquo Medicine andScience in Sports and Exercise vol 38 no 11 pp 1971ndash1979 2006
[13] N A Dimitrova T I Arabadzhiev J-Y Hogrel and G V Dim-itrov ldquoFatigue analysis of interference EMG signals obtainedfrom biceps brachii during isometric voluntary contraction atvarious force levelsrdquo Journal of Electromyography and Kinesiol-ogy vol 19 no 2 pp 252ndash258 2009
[14] D K Kumar S P Arjunan andG R Naik ldquoMeasuring increasein synchronization to identify muscle endurance limitrdquo IEEETransactions on Neural Systems and Rehabilitation Engineeringvol 19 no 5 pp 578ndash587 2011
[15] S Minning C A Eliot T L Uhl and T R Malone ldquoEMGanalysis of shoulder muscle fatigue during resisted isometricshoulder elevationrdquo Journal of Electromyography and Kinesiol-ogy vol 17 no 2 pp 153ndash159 2007
[16] D Farina A Holobar R Merletti and R M Enoka ldquoDecodingthe neural drive to muscles from the surface electromyogramrdquoClinical Neurophysiology vol 121 no 10 pp 1616ndash1623 2010
[17] P Bonato S H Roy M Knaflitz and C J de Luca ldquoTimefrequency parameters of the surface myoelectric signal forassessing muscle fatigue during cyclic dynamic contractionsrdquoIEEE Transactions on Biomedical Engineering vol 48 no 7 pp745ndash753 2001
[18] M McCaffery and A Beebe Pain Clinical Manual for NursingPractice V V Mosby Company Baltimore Md USA 1993
[19] M Gonzalez-Izal A Malanda I Navarro-Amezqueta et alldquoEMG spectral indices and muscle power fatigue duringdynamic contractionsrdquo Journal of Electromyography and Kine-siology vol 20 no 2 pp 233ndash240 2010
[20] G T Allison and T Fujiwara ldquoThe relationship between EMGmedian frequency and low frequency band amplitude changesat different levels of muscle capacityrdquoClinical Biomechanics vol17 no 6 pp 464ndash469 2002
[21] G R Naik D K Kumar V Yadav K Wheeler and S ArjunanldquoTesting of motor unit synchronization model for localizedmuscle fatiguerdquo in Proceedings of the 31st Annual InternationalConference of the IEEE Engineering in Medicine and BiologySociety (EMBS 09) pp 360ndash363 Minneapolis Minn USASeptember 2009
[22] A Cichocki and S-I Amari Adaptive Blind Signal and ImageProcessing John Wiley amp Sons New York NY USA 2002
[23] C D Meyer Matrix Analysis and Applied Linear AlgebraCambridge University Press Cambridge UK 2000
[24] M A Oskoei and H Hu ldquoSupport vector machine-basedclassification scheme for myoelectric control applied to upperlimbrdquo IEEE Transactions on Biomedical Engineering vol 55 no8 pp 1956ndash1965 2008
The aim of this study was to experimentally determinethe most suitable of the currently used sEMG features thatcan be implemented for machine based sEMG analysis toestimate muscle force and fatigue This research has exper-imentally studied six well-accepted features of sEMG andanalyzed the relationship of each of these with force ofmusclecontraction and with muscle fatigue Linear regression andanalysis of variance (ANOVA) were performed to comparethe relationship of each of these features with the force ofmuscle contraction and muscle fatigue The significance ofthis study is that it has shown a comparison between thevarious features of sEMG that have been reported in literatureand has identified the relationship of differences due to threefactors subject force and fatigue
2 Materials and Methods
21 sEMG Recording sEMG signals were recorded using theproprietary Delsys (Boston MA USA) sEMG acquisitionsystem The system supports bipolar recording and has gainof 1000 CMRR of 92 dB and bandwidth of 20ndash450Hz with12 dBoctave roll-off The sampling rate was fixed at 1000samplessecond and the resolutionwas 16 bitssample Activebipolar electrodes (Delsys Boston MA USA) having twosilver bars (1mm wide and 10mm long) mounted directly onthe preamplifier with fixed interelectrode distance of 10mmwere used for recording sEMG
The experiments were conducted on the biceps brachiimuscle because this muscle is superficial and is the primemover for the ankle joint and the muscle fibers run parallelto the surface Two bipolar electrodes were placed above themotor points of the short head of the biceps brachii muscleand inline between the anticubital fossa (depression in thefront of the elbowmdashlateral to the biceps brachii tendon) andthe acromion process (part of the scapula which extends overthe shoulder) at 13rd distance from the anticubital fossaThedistance between the two bipolar electrodes was maintainedat 2 cm Reference electrode was placed on the dorsal sectionand under the elbow Prior to electrode placement the skinarea was cleaned with alcohol swabs and lightly exfoliatedwith paper towel to reduce skin impedance and ensure goodadhesion of the electrodes
22 Experimental Protocol Thirty-five healthy subjects (22male and 13 female aged 22ndash35 years) consented to par-ticipate in these trials The participants with the followingcriteria were excluded (1) arthritis (eg osteoarthritis andrheumatoid arthritis) (2) neuromuscular disorders includingcollagen disorders and nonarticular rheumatism includingfibro myalgia or seizure disorders (3) any recent injuries ofthe hand wrist or arm
The experiments were approved by theUniversityHumanEthics Committee The experiments were performed inaccordance with Declaration of Helsinki of 1975 as revisedin 2004 During the experiments the volunteers were seatedsuch that their feet were flat on the floor the upper arm restedhorizontally on an adjustable desk and the forearm wasvertical (refer to Figure 1) with the elbow at 90 degrees Awall
Force sensor
Force measurement EMG recording unit
Force EMG display
Figure 1 Illustration of the experimental setup
mounted force sensor (S-type force sensormdashINTERFACESM25) was attached to a comfortable hand sized ring with aflexible steel wire padded for comfort and the ring was heldon the palm of the participantThe output of the force sensorwas recorded alongside the sEMG recording on the Delsysacquisition system Before the start of the experiment theexperimental procedures were explained to the participantsand trial runs were performed to familiarize them with theexperiment
Determining Maximum Voluntary Contraction (MVC) Todetermine the maximal voluntary contraction (MVC) threemaximal contractions of 5 seconds were performed with 120seconds rest time between each effortThe participants pulledon the ring and the force of contraction was recorded Theforce output was displayed on the screen and the participantswere encouraged to exert their maximal muscle force andsteadily maintain the force The average of the three readingswas considered to be the MVC If there were any outliers theexperimentwas repeated Preliminary experiments were con-ducted to check the influence of triceps and brachioradialismuscles on biceps during the contraction by recording sEMGof these muscles If the muscle activity of these muscles wasgreater than 5 of MVC of biceps (estimated using RMS ofsEMG) the participants were reseated and assisted in restingthese muscles
After confirming that the triceps and brachii were silentand the activity was concentrated on the biceps muscle theexperiments were conducted during which the participantsperformed three sets of isometric contractions at 50 75and 100 MVC respectively Force of contraction was dis-played on the computer to give them feedback and assistthem in maintaining steady force Participants were asked toperform the contractions as long as they could and until theyexperienced fatigue and pain The trial was terminated whenthe subject was unable to exert the required force and theexerted force dropped below 80 of the target or when thesubjects experienced pain whichever occurred earlier Thepain was subjectively measured using a numeric rating pain
BioMed Research International 3
10
8
6
4
2
0
minus2
minus4
minus6
minus8
0 1 2 3 4 5 6 7 8
Mag
nitu
de (V
)
Time in samples
times10minus5
times104
Figure 2 A sample of the recorded raw sEMG signal (time insamples)
index scale (PIS) [5 18] with range from 0 to 10 and withPIS of 0 corresponding to ldquono painrdquo and 10 corresponding toldquomaximum painrdquo Subjects were informed of the PIS beforethe start of the experiment and were given examples to helpthem understand the scale The subjects were requested andreminded to report the pain only in the biceps muscle Ascore of 8 and above corresponded to the limit of muscleendurance
The total duration of each contraction was referred toas the endurance period This was found to be differentfor different participants and for different levels of musclecontraction To allow a comparison between different exper-iments the time axis was normalized such that the startof the experiment corresponded to T
1and the end of the
exercise was labeled T119864 The participants were given a rest
period of 60 minutes between each contraction but as longas they required The participants were questioned such thatthey reported the muscle to be pain-free and rested Theexperiments performed in this study were similar to ourearlier published protocol [14] A sample raw EMG signalrecorded in the experiment has been shown in Figure 2
23 Data Analysis Data analysis was performed offline onMATLAB 2009a software environment (The MathWorksInc Natick Massachusetts USA) The first step was thetemporal segmentation of the recordings and this was fol-lowed by computation of the six features Regression analysiswas performed to determine the linearity of the relationshipbetween each of the features and force of contraction FinallyN-factor analysis of variance (ANOVA) was computed todetermine the statistical significance of the relationship
231 Temporal Segmentation All sEMG recordings weredivided into 1 second long segments (1000 samples) withoverlap of 100 samples using amovingwindowThe average ofthe feature of all the windows in the corresponding segment
was computed and labeled according to the segment numberThe first segment corresponding to the start of the exercisewas labeled as that of the rested muscle while the finalsegment corresponded to the fatigued muscle at the limit ofendurance
232 Computation of Features The following features werecomputed
(i) Normalized SpectralMoments (NSM5)The spectral fatigueindices (FInsm5) proposed by Dimitrov et al [12] and Dim-itrova et al [13] are the measure of the normalised spectralmoments as follows
FInsm5 =int1198912
1198911
119891minus1sdot PS (119891) sdot 119889119891
int1198912
1198911
1198915 sdot PS (119891) sdot 119889119891 (1)
where PS(f) is the EMG power spectrum 1198911= 8Hz 119891
2=
500Hz and ldquosdotrdquo represents the multiplication factor (see [1213] for further details)
(ii) Median Frequency (MDF) Median frequency is theparticular frequency that divides the power spectrum intotwo sections of equal areas (see [19 20] for details)
(iii) Root Mean Square (RMS) Root mean square (RMS) isthe quadraticmean and a statisticalmeasure of themagnitudeof a time varying signal and is computed using the followingequation
RMS = radic 1119873
119873
sum
119894=1
1199092
119894 (2)
where119873 is the number of samples in the segment and 119909 is thesEMG signal
(iv) Normalised RMS (NRMS) RMS of sEMG for eachparticipant (at 75 and 50 MVC) was normalized withrespect to the RMS of sEMG corresponding to their MVCof the rested muscle and was computed using the followingequation
NRMS =RMS75 50MVC
RMSMVC (3)
(v) Waveform Length (WL) Wavelength is the measure ofthe length of the signal and is computed using the followingequation
WL = 1119873
119873minus1
sum
119894=1
1003816100381610038161003816119909119894+1 minus 1199091198941003816100381610038161003816
(4)
where119873 is the number of samples in the segment and 119909 is thesEMG signal (in samples)
(vi) Increase in Synchronization (IIS) Index Increase in syn-chronization (IIS) index is the measure of independencebetween two signals In this study IIS index was computed
4 BioMed Research International
using the EMG recordings from the two channels (twosensors) [14] using 1 second window length similar to otherfeatures The computation of IIS index has been explainedin detail in our earlier publications [14 21ndash23] The signalwas filtered into four narrow subband components of equalbandwidth (band pass filter with 125Hz frequency band)Independent component analysis (ICA) was performed oneach subband component and the resultant 119899 = 4 unmixingor separating square matrices119882
119899 The global matrix 119866 was
estimated as the product of the 119899th unmixing matrix and theinverse of the (119899 + 1)th unmixing matrix Average of 119866of all the time windows in the segment was computed toobtain 119866 and IIS index corresponding to the segment wasobtained by computing log 119866 the normalized determinantof the global matrix
233 Regression Analysis To determine the relationshipbetween the level of contraction (force) and the featuresregression analysis was performed For this analysis thefeatures computed from the sEMG were recorded during theinitial state of the muscle contraction Preliminary analysisshowed that the relationships were linear in nature Basedon this criterion the relationship between the level of con-traction during initial state of maintaining the force and thefeatures was computed using linear regression analysis with95 confidence intervals
234 Statistical Analysis The statistical significance of theeffect and the relationship between the different factors oneach of the six features of sEMG was studied Three-wayanalysis of variance (ANOVA) with interaction conductedfor each of the 6 features with 95 confidence interval (119875 lt005) was performed Kurtosis measures and skew tests wereperformed to check and confirm the underlying assumptionsof ANOVA in analysing the data The three factors were(i) difference between subjects (ii) level of contraction (MVC) and (iii) muscle fatigue In this study ldquolocalisedmuscle fatiguerdquo was defined as the conditionwhen the subjectwas unable to perform the level of contraction and reporteda high level of discomfort (gt8 on PIS)
The factor subject was considered as a random factorANOVA model was designed to identify the significance ofthe different factors on each of the six features of sEMG Theresult of this analysis would indicate the features that are bestable to identify the effect due to the factor and also determinethe strength of the relationship between the different factors
3 Results
Table 1 is the summary of the results from the three factormain effects using the ANOVA model repeated for each ofthe six features The three factors considered were (i) subject(ii) level of contraction ( MVC) and (iii) localised musclefatigue The effect of localised muscle fatigue was definedas the difference between near the start and near the endof the exercise From Table 1 it is observed that the level ofcontraction ( MVC) as a main effect had significant effectonNSM5 (119875 lt 001) RMS (119875 lt 005) NRMS (119875 lt 005) and
Table 1 Three-way ANOVA for each of the six features with95 confidence interval The three factors and their consideredinteractions are subject level of contraction ( MVC) and musclefatigue factor (initial and final segments of the exercise) forisometric contraction and for the six features of sEMG
FeaturesEffect ofsubject119865-value (119875)
Effect of level ofcontraction ( MVC)119865-value (119875)
IIS index (119875 lt 005) From this table it is also observed thatthe statistical significant relationship of muscle fatigue wasonly with IIS index of sEMG (119875 lt 001) The effect on NMRSat 100 MVC was ignored for the statistical and regressionanalysis because this was the basis for normalisation of therest of the data
Table 2 shows the results of the regression analysis per-formed using the features of sEMG computed from the initialstate of the muscle contraction From this table it is observedthat the relationship of NSM5 of sEMG with muscle forcewas the most linear in comparison with other features (1198772 =096) NRMS also showed a good linearity relationship with1198772 value of 093 The bar plots showing the mean values
(and standard deviation) of the different features with level ofcontraction ( of MVC) are shown in Figure 3 These plotsconfirm the observations from Tables 1 and 2 that NSM5 hasa strong linear relationship with force of muscle contraction
4 Discussion and Conclusion
The association between the sEMG and force of musclecontraction is well accepted Different features of sEMG suchas RMS median frequency normalized spectral momentsand wavelength [7ndash9 11ndash13] have been studied in relationwith the muscle force However none of these studies havereported a comparison between the different features todetermine the statistical significance of these relationshipsnor have these analyzed the linearity of these relationships
BioMed Research International 5
0
50
100
150M
DF
50 75 100Level of contraction (MVC)
(a)
50 75 1000
1
2
3
4
RMS
Level of contraction (MVC)
(b)
0123456
NSM
50 75 100Level of contraction (MVC)
times104
(c)
50 75 100
0
02
IIS
Level of contraction (MVC)
minus06
minus04
minus02
(d)
50 75 1000
05
1
15
Level of contraction (MVC)
NRM
S
(e)
0001002003004
WL
50 75 100Level of contraction (MVC)
minus002
minus001
(f)
Figure 3 Average (plusmn SD) of six features of sEMG and level of contraction for the rested muscle (a) MDF (b) RMS (c) normalised spectralindex (NSM5) (d) IIS index (e) normalised RMS and (f) WL
Such an analysis is important for direct interpretation ofsEMG to estimate the force of muscle contraction
This study has experimentally studied and compared therelationship of muscle force and muscle fatigue with each ofthe six well-accepted features of sEMG Analysis of variance(ANOVA) was conducted to determine the effect of threemain different factors level of muscle contraction ( MVC)muscle fatigue and intersubject variation ANOVA resultsshow that RMS (119875 lt 005) and NSM5 (119875 lt 001) of sEMGwere significantly affected by the level of muscle contraction( MVC) The results also indicate that IIS index (119875 lt 001)was significant in identifying the muscle endurance limit
It is observed from the results that the intersubjectvariation was significant for RMS NRMS andWL while thevariation was not significant for NSM5 and IIS index (119875 gt005) Features such as RMS NRMS and WL are associatedwith the amplitude of the signal and the results indicate thatthere is a significant variation in the amplitude of sEMGbetween subjects However other features that are based onspectrum such as NSM5 or entropy dependent such as IISindex do not have significant differences between subjects
The linearity analysis indicates that the most linearrelationship of force of contraction (ranging from 50 to 100
MVC) was with NSM5 (1198772 = 096) followed by normalizedRMS (1198772 = 093) The linearity relationship between RMSof the signal and force of contraction was poor (1198772 = 071)This was also observed from the plots (Figure 2) The resultssuggest that the relationship between NSM5 a normalizedmeasure of the spectrumof the signal [12] and force ofmusclecontraction ( MVC) is the most significant and linear Theresults also indicate that for biceps NSM5 do not requireany normalization or calibrationThis indicates that NSM5 issuitable to estimate the level of muscle contraction comparedwith other features [6 24] even though this has not beenreported in literature
While earlier studies have identified NSM5 to be a fatigueindex [12 13] this study has shown that NSM5 of sEMGis the measure of force of muscle contraction This maybe attributed to the spectrum of sEMG being significantlyinfluenced by the rate of muscle stimulation and thus withforce of muscle contraction This study has also confirmedthat IIS index [14] is the most suitable indicator of musclebeing at the limit of endurance and is fatiguedThe study alsofound that the advantage of bothNSM5 and IIS was that thesedid not require any normalization
6 BioMed Research International
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
References
[1] J V Basmajian and C J de LucaMuscles Alive Their FunctionsRevealed by Electromyography Williams and Wilkins Balti-more Md USA 1985
[2] R Merletti and S Roy ldquoMyoelectric and mechanical manifes-tations of muscle fatigue in voluntary contractionsrdquo Journal ofOrthopaedic and Sports PhysicalTherapy vol 24 no 6 pp 342ndash353 1996
[3] T J Dartnall M A Nordstrom and J G Semmler ldquoMotor unitsynchronization is increased in biceps brachii after exercise-induced damage to elbow flexormusclesrdquo Journal of Neurophys-iology vol 99 no 2 pp 1008ndash1019 2008
[4] F E Marino M Gard and E J Drinkwater ldquoThe limits toexercise performance and the future of fatigue researchrdquo BritishJournal of Sports Medicine vol 45 no 1 pp 65ndash67 2011
[5] F A BainbridgeThePhysiology ofMuscular Exercise LongmanGreen and Co New York NY USA 3rd edition 1931
[6] S G Boe C L Rice and T J Doherty ldquoEstimating contractionlevel using root mean square amplitude in control subjects andpatients with neuromuscular disordersrdquo Archives of PhysicalMedicine and Rehabilitation vol 89 no 4 pp 711ndash718 2008
[7] RMerletti and P Parker Electromyography JohnWiley amp SonsNew York NY USA 2004
[8] D K Kumar N D Pah and A Bradley ldquoWavelet analysis ofsurface electromyography to determine muscle fatiguerdquo IEEETransactions on Neural Systems and Rehabilitation Engineeringvol 11 no 4 pp 400ndash406 2003
[9] D Moshou I Hostens G Papaioannou and H RamonldquoDynamicmuscle fatigue detection using self-organizingmapsrdquoApplied Soft Computing Journal vol 5 no 4 pp 391ndash398 2005
[10] S P Arjunan and D K Kumar ldquoDecoding subtle forearmflexions using fractal features of surface electromyogram fromsingle and multiple sensorsrdquo Journal of NeuroEngineering andRehabilitation vol 7 no 1 article 53 2010
[11] G Wang X-M Ren L Li and Z-Z Wang ldquoMultifractalanalysis of surface EMG signals for assessing muscle fatigueduring static contractionsrdquo Journal of Zhejiang University AScience vol 8 no 6 pp 910ndash915 2007
[12] G V Dimitrov T I Arabadzhiev K N Mileva J L Bowtell NCrichton andNADimitrova ldquoMuscle fatigue during dynamiccontractions assessed by new spectral indicesrdquo Medicine andScience in Sports and Exercise vol 38 no 11 pp 1971ndash1979 2006
[13] N A Dimitrova T I Arabadzhiev J-Y Hogrel and G V Dim-itrov ldquoFatigue analysis of interference EMG signals obtainedfrom biceps brachii during isometric voluntary contraction atvarious force levelsrdquo Journal of Electromyography and Kinesiol-ogy vol 19 no 2 pp 252ndash258 2009
[14] D K Kumar S P Arjunan andG R Naik ldquoMeasuring increasein synchronization to identify muscle endurance limitrdquo IEEETransactions on Neural Systems and Rehabilitation Engineeringvol 19 no 5 pp 578ndash587 2011
[15] S Minning C A Eliot T L Uhl and T R Malone ldquoEMGanalysis of shoulder muscle fatigue during resisted isometricshoulder elevationrdquo Journal of Electromyography and Kinesiol-ogy vol 17 no 2 pp 153ndash159 2007
[16] D Farina A Holobar R Merletti and R M Enoka ldquoDecodingthe neural drive to muscles from the surface electromyogramrdquoClinical Neurophysiology vol 121 no 10 pp 1616ndash1623 2010
[17] P Bonato S H Roy M Knaflitz and C J de Luca ldquoTimefrequency parameters of the surface myoelectric signal forassessing muscle fatigue during cyclic dynamic contractionsrdquoIEEE Transactions on Biomedical Engineering vol 48 no 7 pp745ndash753 2001
[18] M McCaffery and A Beebe Pain Clinical Manual for NursingPractice V V Mosby Company Baltimore Md USA 1993
[19] M Gonzalez-Izal A Malanda I Navarro-Amezqueta et alldquoEMG spectral indices and muscle power fatigue duringdynamic contractionsrdquo Journal of Electromyography and Kine-siology vol 20 no 2 pp 233ndash240 2010
[20] G T Allison and T Fujiwara ldquoThe relationship between EMGmedian frequency and low frequency band amplitude changesat different levels of muscle capacityrdquoClinical Biomechanics vol17 no 6 pp 464ndash469 2002
[21] G R Naik D K Kumar V Yadav K Wheeler and S ArjunanldquoTesting of motor unit synchronization model for localizedmuscle fatiguerdquo in Proceedings of the 31st Annual InternationalConference of the IEEE Engineering in Medicine and BiologySociety (EMBS 09) pp 360ndash363 Minneapolis Minn USASeptember 2009
[22] A Cichocki and S-I Amari Adaptive Blind Signal and ImageProcessing John Wiley amp Sons New York NY USA 2002
[23] C D Meyer Matrix Analysis and Applied Linear AlgebraCambridge University Press Cambridge UK 2000
[24] M A Oskoei and H Hu ldquoSupport vector machine-basedclassification scheme for myoelectric control applied to upperlimbrdquo IEEE Transactions on Biomedical Engineering vol 55 no8 pp 1956ndash1965 2008
Figure 2 A sample of the recorded raw sEMG signal (time insamples)
index scale (PIS) [5 18] with range from 0 to 10 and withPIS of 0 corresponding to ldquono painrdquo and 10 corresponding toldquomaximum painrdquo Subjects were informed of the PIS beforethe start of the experiment and were given examples to helpthem understand the scale The subjects were requested andreminded to report the pain only in the biceps muscle Ascore of 8 and above corresponded to the limit of muscleendurance
The total duration of each contraction was referred toas the endurance period This was found to be differentfor different participants and for different levels of musclecontraction To allow a comparison between different exper-iments the time axis was normalized such that the startof the experiment corresponded to T
1and the end of the
exercise was labeled T119864 The participants were given a rest
period of 60 minutes between each contraction but as longas they required The participants were questioned such thatthey reported the muscle to be pain-free and rested Theexperiments performed in this study were similar to ourearlier published protocol [14] A sample raw EMG signalrecorded in the experiment has been shown in Figure 2
23 Data Analysis Data analysis was performed offline onMATLAB 2009a software environment (The MathWorksInc Natick Massachusetts USA) The first step was thetemporal segmentation of the recordings and this was fol-lowed by computation of the six features Regression analysiswas performed to determine the linearity of the relationshipbetween each of the features and force of contraction FinallyN-factor analysis of variance (ANOVA) was computed todetermine the statistical significance of the relationship
231 Temporal Segmentation All sEMG recordings weredivided into 1 second long segments (1000 samples) withoverlap of 100 samples using amovingwindowThe average ofthe feature of all the windows in the corresponding segment
was computed and labeled according to the segment numberThe first segment corresponding to the start of the exercisewas labeled as that of the rested muscle while the finalsegment corresponded to the fatigued muscle at the limit ofendurance
232 Computation of Features The following features werecomputed
(i) Normalized SpectralMoments (NSM5)The spectral fatigueindices (FInsm5) proposed by Dimitrov et al [12] and Dim-itrova et al [13] are the measure of the normalised spectralmoments as follows
FInsm5 =int1198912
1198911
119891minus1sdot PS (119891) sdot 119889119891
int1198912
1198911
1198915 sdot PS (119891) sdot 119889119891 (1)
where PS(f) is the EMG power spectrum 1198911= 8Hz 119891
2=
500Hz and ldquosdotrdquo represents the multiplication factor (see [1213] for further details)
(ii) Median Frequency (MDF) Median frequency is theparticular frequency that divides the power spectrum intotwo sections of equal areas (see [19 20] for details)
(iii) Root Mean Square (RMS) Root mean square (RMS) isthe quadraticmean and a statisticalmeasure of themagnitudeof a time varying signal and is computed using the followingequation
RMS = radic 1119873
119873
sum
119894=1
1199092
119894 (2)
where119873 is the number of samples in the segment and 119909 is thesEMG signal
(iv) Normalised RMS (NRMS) RMS of sEMG for eachparticipant (at 75 and 50 MVC) was normalized withrespect to the RMS of sEMG corresponding to their MVCof the rested muscle and was computed using the followingequation
NRMS =RMS75 50MVC
RMSMVC (3)
(v) Waveform Length (WL) Wavelength is the measure ofthe length of the signal and is computed using the followingequation
WL = 1119873
119873minus1
sum
119894=1
1003816100381610038161003816119909119894+1 minus 1199091198941003816100381610038161003816
(4)
where119873 is the number of samples in the segment and 119909 is thesEMG signal (in samples)
(vi) Increase in Synchronization (IIS) Index Increase in syn-chronization (IIS) index is the measure of independencebetween two signals In this study IIS index was computed
4 BioMed Research International
using the EMG recordings from the two channels (twosensors) [14] using 1 second window length similar to otherfeatures The computation of IIS index has been explainedin detail in our earlier publications [14 21ndash23] The signalwas filtered into four narrow subband components of equalbandwidth (band pass filter with 125Hz frequency band)Independent component analysis (ICA) was performed oneach subband component and the resultant 119899 = 4 unmixingor separating square matrices119882
119899 The global matrix 119866 was
estimated as the product of the 119899th unmixing matrix and theinverse of the (119899 + 1)th unmixing matrix Average of 119866of all the time windows in the segment was computed toobtain 119866 and IIS index corresponding to the segment wasobtained by computing log 119866 the normalized determinantof the global matrix
233 Regression Analysis To determine the relationshipbetween the level of contraction (force) and the featuresregression analysis was performed For this analysis thefeatures computed from the sEMG were recorded during theinitial state of the muscle contraction Preliminary analysisshowed that the relationships were linear in nature Basedon this criterion the relationship between the level of con-traction during initial state of maintaining the force and thefeatures was computed using linear regression analysis with95 confidence intervals
234 Statistical Analysis The statistical significance of theeffect and the relationship between the different factors oneach of the six features of sEMG was studied Three-wayanalysis of variance (ANOVA) with interaction conductedfor each of the 6 features with 95 confidence interval (119875 lt005) was performed Kurtosis measures and skew tests wereperformed to check and confirm the underlying assumptionsof ANOVA in analysing the data The three factors were(i) difference between subjects (ii) level of contraction (MVC) and (iii) muscle fatigue In this study ldquolocalisedmuscle fatiguerdquo was defined as the conditionwhen the subjectwas unable to perform the level of contraction and reporteda high level of discomfort (gt8 on PIS)
The factor subject was considered as a random factorANOVA model was designed to identify the significance ofthe different factors on each of the six features of sEMG Theresult of this analysis would indicate the features that are bestable to identify the effect due to the factor and also determinethe strength of the relationship between the different factors
3 Results
Table 1 is the summary of the results from the three factormain effects using the ANOVA model repeated for each ofthe six features The three factors considered were (i) subject(ii) level of contraction ( MVC) and (iii) localised musclefatigue The effect of localised muscle fatigue was definedas the difference between near the start and near the endof the exercise From Table 1 it is observed that the level ofcontraction ( MVC) as a main effect had significant effectonNSM5 (119875 lt 001) RMS (119875 lt 005) NRMS (119875 lt 005) and
Table 1 Three-way ANOVA for each of the six features with95 confidence interval The three factors and their consideredinteractions are subject level of contraction ( MVC) and musclefatigue factor (initial and final segments of the exercise) forisometric contraction and for the six features of sEMG
FeaturesEffect ofsubject119865-value (119875)
Effect of level ofcontraction ( MVC)119865-value (119875)
IIS index (119875 lt 005) From this table it is also observed thatthe statistical significant relationship of muscle fatigue wasonly with IIS index of sEMG (119875 lt 001) The effect on NMRSat 100 MVC was ignored for the statistical and regressionanalysis because this was the basis for normalisation of therest of the data
Table 2 shows the results of the regression analysis per-formed using the features of sEMG computed from the initialstate of the muscle contraction From this table it is observedthat the relationship of NSM5 of sEMG with muscle forcewas the most linear in comparison with other features (1198772 =096) NRMS also showed a good linearity relationship with1198772 value of 093 The bar plots showing the mean values
(and standard deviation) of the different features with level ofcontraction ( of MVC) are shown in Figure 3 These plotsconfirm the observations from Tables 1 and 2 that NSM5 hasa strong linear relationship with force of muscle contraction
4 Discussion and Conclusion
The association between the sEMG and force of musclecontraction is well accepted Different features of sEMG suchas RMS median frequency normalized spectral momentsand wavelength [7ndash9 11ndash13] have been studied in relationwith the muscle force However none of these studies havereported a comparison between the different features todetermine the statistical significance of these relationshipsnor have these analyzed the linearity of these relationships
BioMed Research International 5
0
50
100
150M
DF
50 75 100Level of contraction (MVC)
(a)
50 75 1000
1
2
3
4
RMS
Level of contraction (MVC)
(b)
0123456
NSM
50 75 100Level of contraction (MVC)
times104
(c)
50 75 100
0
02
IIS
Level of contraction (MVC)
minus06
minus04
minus02
(d)
50 75 1000
05
1
15
Level of contraction (MVC)
NRM
S
(e)
0001002003004
WL
50 75 100Level of contraction (MVC)
minus002
minus001
(f)
Figure 3 Average (plusmn SD) of six features of sEMG and level of contraction for the rested muscle (a) MDF (b) RMS (c) normalised spectralindex (NSM5) (d) IIS index (e) normalised RMS and (f) WL
Such an analysis is important for direct interpretation ofsEMG to estimate the force of muscle contraction
This study has experimentally studied and compared therelationship of muscle force and muscle fatigue with each ofthe six well-accepted features of sEMG Analysis of variance(ANOVA) was conducted to determine the effect of threemain different factors level of muscle contraction ( MVC)muscle fatigue and intersubject variation ANOVA resultsshow that RMS (119875 lt 005) and NSM5 (119875 lt 001) of sEMGwere significantly affected by the level of muscle contraction( MVC) The results also indicate that IIS index (119875 lt 001)was significant in identifying the muscle endurance limit
It is observed from the results that the intersubjectvariation was significant for RMS NRMS andWL while thevariation was not significant for NSM5 and IIS index (119875 gt005) Features such as RMS NRMS and WL are associatedwith the amplitude of the signal and the results indicate thatthere is a significant variation in the amplitude of sEMGbetween subjects However other features that are based onspectrum such as NSM5 or entropy dependent such as IISindex do not have significant differences between subjects
The linearity analysis indicates that the most linearrelationship of force of contraction (ranging from 50 to 100
MVC) was with NSM5 (1198772 = 096) followed by normalizedRMS (1198772 = 093) The linearity relationship between RMSof the signal and force of contraction was poor (1198772 = 071)This was also observed from the plots (Figure 2) The resultssuggest that the relationship between NSM5 a normalizedmeasure of the spectrumof the signal [12] and force ofmusclecontraction ( MVC) is the most significant and linear Theresults also indicate that for biceps NSM5 do not requireany normalization or calibrationThis indicates that NSM5 issuitable to estimate the level of muscle contraction comparedwith other features [6 24] even though this has not beenreported in literature
While earlier studies have identified NSM5 to be a fatigueindex [12 13] this study has shown that NSM5 of sEMGis the measure of force of muscle contraction This maybe attributed to the spectrum of sEMG being significantlyinfluenced by the rate of muscle stimulation and thus withforce of muscle contraction This study has also confirmedthat IIS index [14] is the most suitable indicator of musclebeing at the limit of endurance and is fatiguedThe study alsofound that the advantage of bothNSM5 and IIS was that thesedid not require any normalization
6 BioMed Research International
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
References
[1] J V Basmajian and C J de LucaMuscles Alive Their FunctionsRevealed by Electromyography Williams and Wilkins Balti-more Md USA 1985
[2] R Merletti and S Roy ldquoMyoelectric and mechanical manifes-tations of muscle fatigue in voluntary contractionsrdquo Journal ofOrthopaedic and Sports PhysicalTherapy vol 24 no 6 pp 342ndash353 1996
[3] T J Dartnall M A Nordstrom and J G Semmler ldquoMotor unitsynchronization is increased in biceps brachii after exercise-induced damage to elbow flexormusclesrdquo Journal of Neurophys-iology vol 99 no 2 pp 1008ndash1019 2008
[4] F E Marino M Gard and E J Drinkwater ldquoThe limits toexercise performance and the future of fatigue researchrdquo BritishJournal of Sports Medicine vol 45 no 1 pp 65ndash67 2011
[5] F A BainbridgeThePhysiology ofMuscular Exercise LongmanGreen and Co New York NY USA 3rd edition 1931
[6] S G Boe C L Rice and T J Doherty ldquoEstimating contractionlevel using root mean square amplitude in control subjects andpatients with neuromuscular disordersrdquo Archives of PhysicalMedicine and Rehabilitation vol 89 no 4 pp 711ndash718 2008
[7] RMerletti and P Parker Electromyography JohnWiley amp SonsNew York NY USA 2004
[8] D K Kumar N D Pah and A Bradley ldquoWavelet analysis ofsurface electromyography to determine muscle fatiguerdquo IEEETransactions on Neural Systems and Rehabilitation Engineeringvol 11 no 4 pp 400ndash406 2003
[9] D Moshou I Hostens G Papaioannou and H RamonldquoDynamicmuscle fatigue detection using self-organizingmapsrdquoApplied Soft Computing Journal vol 5 no 4 pp 391ndash398 2005
[10] S P Arjunan and D K Kumar ldquoDecoding subtle forearmflexions using fractal features of surface electromyogram fromsingle and multiple sensorsrdquo Journal of NeuroEngineering andRehabilitation vol 7 no 1 article 53 2010
[11] G Wang X-M Ren L Li and Z-Z Wang ldquoMultifractalanalysis of surface EMG signals for assessing muscle fatigueduring static contractionsrdquo Journal of Zhejiang University AScience vol 8 no 6 pp 910ndash915 2007
[12] G V Dimitrov T I Arabadzhiev K N Mileva J L Bowtell NCrichton andNADimitrova ldquoMuscle fatigue during dynamiccontractions assessed by new spectral indicesrdquo Medicine andScience in Sports and Exercise vol 38 no 11 pp 1971ndash1979 2006
[13] N A Dimitrova T I Arabadzhiev J-Y Hogrel and G V Dim-itrov ldquoFatigue analysis of interference EMG signals obtainedfrom biceps brachii during isometric voluntary contraction atvarious force levelsrdquo Journal of Electromyography and Kinesiol-ogy vol 19 no 2 pp 252ndash258 2009
[14] D K Kumar S P Arjunan andG R Naik ldquoMeasuring increasein synchronization to identify muscle endurance limitrdquo IEEETransactions on Neural Systems and Rehabilitation Engineeringvol 19 no 5 pp 578ndash587 2011
[15] S Minning C A Eliot T L Uhl and T R Malone ldquoEMGanalysis of shoulder muscle fatigue during resisted isometricshoulder elevationrdquo Journal of Electromyography and Kinesiol-ogy vol 17 no 2 pp 153ndash159 2007
[16] D Farina A Holobar R Merletti and R M Enoka ldquoDecodingthe neural drive to muscles from the surface electromyogramrdquoClinical Neurophysiology vol 121 no 10 pp 1616ndash1623 2010
[17] P Bonato S H Roy M Knaflitz and C J de Luca ldquoTimefrequency parameters of the surface myoelectric signal forassessing muscle fatigue during cyclic dynamic contractionsrdquoIEEE Transactions on Biomedical Engineering vol 48 no 7 pp745ndash753 2001
[18] M McCaffery and A Beebe Pain Clinical Manual for NursingPractice V V Mosby Company Baltimore Md USA 1993
[19] M Gonzalez-Izal A Malanda I Navarro-Amezqueta et alldquoEMG spectral indices and muscle power fatigue duringdynamic contractionsrdquo Journal of Electromyography and Kine-siology vol 20 no 2 pp 233ndash240 2010
[20] G T Allison and T Fujiwara ldquoThe relationship between EMGmedian frequency and low frequency band amplitude changesat different levels of muscle capacityrdquoClinical Biomechanics vol17 no 6 pp 464ndash469 2002
[21] G R Naik D K Kumar V Yadav K Wheeler and S ArjunanldquoTesting of motor unit synchronization model for localizedmuscle fatiguerdquo in Proceedings of the 31st Annual InternationalConference of the IEEE Engineering in Medicine and BiologySociety (EMBS 09) pp 360ndash363 Minneapolis Minn USASeptember 2009
[22] A Cichocki and S-I Amari Adaptive Blind Signal and ImageProcessing John Wiley amp Sons New York NY USA 2002
[23] C D Meyer Matrix Analysis and Applied Linear AlgebraCambridge University Press Cambridge UK 2000
[24] M A Oskoei and H Hu ldquoSupport vector machine-basedclassification scheme for myoelectric control applied to upperlimbrdquo IEEE Transactions on Biomedical Engineering vol 55 no8 pp 1956ndash1965 2008
using the EMG recordings from the two channels (twosensors) [14] using 1 second window length similar to otherfeatures The computation of IIS index has been explainedin detail in our earlier publications [14 21ndash23] The signalwas filtered into four narrow subband components of equalbandwidth (band pass filter with 125Hz frequency band)Independent component analysis (ICA) was performed oneach subband component and the resultant 119899 = 4 unmixingor separating square matrices119882
119899 The global matrix 119866 was
estimated as the product of the 119899th unmixing matrix and theinverse of the (119899 + 1)th unmixing matrix Average of 119866of all the time windows in the segment was computed toobtain 119866 and IIS index corresponding to the segment wasobtained by computing log 119866 the normalized determinantof the global matrix
233 Regression Analysis To determine the relationshipbetween the level of contraction (force) and the featuresregression analysis was performed For this analysis thefeatures computed from the sEMG were recorded during theinitial state of the muscle contraction Preliminary analysisshowed that the relationships were linear in nature Basedon this criterion the relationship between the level of con-traction during initial state of maintaining the force and thefeatures was computed using linear regression analysis with95 confidence intervals
234 Statistical Analysis The statistical significance of theeffect and the relationship between the different factors oneach of the six features of sEMG was studied Three-wayanalysis of variance (ANOVA) with interaction conductedfor each of the 6 features with 95 confidence interval (119875 lt005) was performed Kurtosis measures and skew tests wereperformed to check and confirm the underlying assumptionsof ANOVA in analysing the data The three factors were(i) difference between subjects (ii) level of contraction (MVC) and (iii) muscle fatigue In this study ldquolocalisedmuscle fatiguerdquo was defined as the conditionwhen the subjectwas unable to perform the level of contraction and reporteda high level of discomfort (gt8 on PIS)
The factor subject was considered as a random factorANOVA model was designed to identify the significance ofthe different factors on each of the six features of sEMG Theresult of this analysis would indicate the features that are bestable to identify the effect due to the factor and also determinethe strength of the relationship between the different factors
3 Results
Table 1 is the summary of the results from the three factormain effects using the ANOVA model repeated for each ofthe six features The three factors considered were (i) subject(ii) level of contraction ( MVC) and (iii) localised musclefatigue The effect of localised muscle fatigue was definedas the difference between near the start and near the endof the exercise From Table 1 it is observed that the level ofcontraction ( MVC) as a main effect had significant effectonNSM5 (119875 lt 001) RMS (119875 lt 005) NRMS (119875 lt 005) and
Table 1 Three-way ANOVA for each of the six features with95 confidence interval The three factors and their consideredinteractions are subject level of contraction ( MVC) and musclefatigue factor (initial and final segments of the exercise) forisometric contraction and for the six features of sEMG
FeaturesEffect ofsubject119865-value (119875)
Effect of level ofcontraction ( MVC)119865-value (119875)
IIS index (119875 lt 005) From this table it is also observed thatthe statistical significant relationship of muscle fatigue wasonly with IIS index of sEMG (119875 lt 001) The effect on NMRSat 100 MVC was ignored for the statistical and regressionanalysis because this was the basis for normalisation of therest of the data
Table 2 shows the results of the regression analysis per-formed using the features of sEMG computed from the initialstate of the muscle contraction From this table it is observedthat the relationship of NSM5 of sEMG with muscle forcewas the most linear in comparison with other features (1198772 =096) NRMS also showed a good linearity relationship with1198772 value of 093 The bar plots showing the mean values
(and standard deviation) of the different features with level ofcontraction ( of MVC) are shown in Figure 3 These plotsconfirm the observations from Tables 1 and 2 that NSM5 hasa strong linear relationship with force of muscle contraction
4 Discussion and Conclusion
The association between the sEMG and force of musclecontraction is well accepted Different features of sEMG suchas RMS median frequency normalized spectral momentsand wavelength [7ndash9 11ndash13] have been studied in relationwith the muscle force However none of these studies havereported a comparison between the different features todetermine the statistical significance of these relationshipsnor have these analyzed the linearity of these relationships
BioMed Research International 5
0
50
100
150M
DF
50 75 100Level of contraction (MVC)
(a)
50 75 1000
1
2
3
4
RMS
Level of contraction (MVC)
(b)
0123456
NSM
50 75 100Level of contraction (MVC)
times104
(c)
50 75 100
0
02
IIS
Level of contraction (MVC)
minus06
minus04
minus02
(d)
50 75 1000
05
1
15
Level of contraction (MVC)
NRM
S
(e)
0001002003004
WL
50 75 100Level of contraction (MVC)
minus002
minus001
(f)
Figure 3 Average (plusmn SD) of six features of sEMG and level of contraction for the rested muscle (a) MDF (b) RMS (c) normalised spectralindex (NSM5) (d) IIS index (e) normalised RMS and (f) WL
Such an analysis is important for direct interpretation ofsEMG to estimate the force of muscle contraction
This study has experimentally studied and compared therelationship of muscle force and muscle fatigue with each ofthe six well-accepted features of sEMG Analysis of variance(ANOVA) was conducted to determine the effect of threemain different factors level of muscle contraction ( MVC)muscle fatigue and intersubject variation ANOVA resultsshow that RMS (119875 lt 005) and NSM5 (119875 lt 001) of sEMGwere significantly affected by the level of muscle contraction( MVC) The results also indicate that IIS index (119875 lt 001)was significant in identifying the muscle endurance limit
It is observed from the results that the intersubjectvariation was significant for RMS NRMS andWL while thevariation was not significant for NSM5 and IIS index (119875 gt005) Features such as RMS NRMS and WL are associatedwith the amplitude of the signal and the results indicate thatthere is a significant variation in the amplitude of sEMGbetween subjects However other features that are based onspectrum such as NSM5 or entropy dependent such as IISindex do not have significant differences between subjects
The linearity analysis indicates that the most linearrelationship of force of contraction (ranging from 50 to 100
MVC) was with NSM5 (1198772 = 096) followed by normalizedRMS (1198772 = 093) The linearity relationship between RMSof the signal and force of contraction was poor (1198772 = 071)This was also observed from the plots (Figure 2) The resultssuggest that the relationship between NSM5 a normalizedmeasure of the spectrumof the signal [12] and force ofmusclecontraction ( MVC) is the most significant and linear Theresults also indicate that for biceps NSM5 do not requireany normalization or calibrationThis indicates that NSM5 issuitable to estimate the level of muscle contraction comparedwith other features [6 24] even though this has not beenreported in literature
While earlier studies have identified NSM5 to be a fatigueindex [12 13] this study has shown that NSM5 of sEMGis the measure of force of muscle contraction This maybe attributed to the spectrum of sEMG being significantlyinfluenced by the rate of muscle stimulation and thus withforce of muscle contraction This study has also confirmedthat IIS index [14] is the most suitable indicator of musclebeing at the limit of endurance and is fatiguedThe study alsofound that the advantage of bothNSM5 and IIS was that thesedid not require any normalization
6 BioMed Research International
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
References
[1] J V Basmajian and C J de LucaMuscles Alive Their FunctionsRevealed by Electromyography Williams and Wilkins Balti-more Md USA 1985
[2] R Merletti and S Roy ldquoMyoelectric and mechanical manifes-tations of muscle fatigue in voluntary contractionsrdquo Journal ofOrthopaedic and Sports PhysicalTherapy vol 24 no 6 pp 342ndash353 1996
[3] T J Dartnall M A Nordstrom and J G Semmler ldquoMotor unitsynchronization is increased in biceps brachii after exercise-induced damage to elbow flexormusclesrdquo Journal of Neurophys-iology vol 99 no 2 pp 1008ndash1019 2008
[4] F E Marino M Gard and E J Drinkwater ldquoThe limits toexercise performance and the future of fatigue researchrdquo BritishJournal of Sports Medicine vol 45 no 1 pp 65ndash67 2011
[5] F A BainbridgeThePhysiology ofMuscular Exercise LongmanGreen and Co New York NY USA 3rd edition 1931
[6] S G Boe C L Rice and T J Doherty ldquoEstimating contractionlevel using root mean square amplitude in control subjects andpatients with neuromuscular disordersrdquo Archives of PhysicalMedicine and Rehabilitation vol 89 no 4 pp 711ndash718 2008
[7] RMerletti and P Parker Electromyography JohnWiley amp SonsNew York NY USA 2004
[8] D K Kumar N D Pah and A Bradley ldquoWavelet analysis ofsurface electromyography to determine muscle fatiguerdquo IEEETransactions on Neural Systems and Rehabilitation Engineeringvol 11 no 4 pp 400ndash406 2003
[9] D Moshou I Hostens G Papaioannou and H RamonldquoDynamicmuscle fatigue detection using self-organizingmapsrdquoApplied Soft Computing Journal vol 5 no 4 pp 391ndash398 2005
[10] S P Arjunan and D K Kumar ldquoDecoding subtle forearmflexions using fractal features of surface electromyogram fromsingle and multiple sensorsrdquo Journal of NeuroEngineering andRehabilitation vol 7 no 1 article 53 2010
[11] G Wang X-M Ren L Li and Z-Z Wang ldquoMultifractalanalysis of surface EMG signals for assessing muscle fatigueduring static contractionsrdquo Journal of Zhejiang University AScience vol 8 no 6 pp 910ndash915 2007
[12] G V Dimitrov T I Arabadzhiev K N Mileva J L Bowtell NCrichton andNADimitrova ldquoMuscle fatigue during dynamiccontractions assessed by new spectral indicesrdquo Medicine andScience in Sports and Exercise vol 38 no 11 pp 1971ndash1979 2006
[13] N A Dimitrova T I Arabadzhiev J-Y Hogrel and G V Dim-itrov ldquoFatigue analysis of interference EMG signals obtainedfrom biceps brachii during isometric voluntary contraction atvarious force levelsrdquo Journal of Electromyography and Kinesiol-ogy vol 19 no 2 pp 252ndash258 2009
[14] D K Kumar S P Arjunan andG R Naik ldquoMeasuring increasein synchronization to identify muscle endurance limitrdquo IEEETransactions on Neural Systems and Rehabilitation Engineeringvol 19 no 5 pp 578ndash587 2011
[15] S Minning C A Eliot T L Uhl and T R Malone ldquoEMGanalysis of shoulder muscle fatigue during resisted isometricshoulder elevationrdquo Journal of Electromyography and Kinesiol-ogy vol 17 no 2 pp 153ndash159 2007
[16] D Farina A Holobar R Merletti and R M Enoka ldquoDecodingthe neural drive to muscles from the surface electromyogramrdquoClinical Neurophysiology vol 121 no 10 pp 1616ndash1623 2010
[17] P Bonato S H Roy M Knaflitz and C J de Luca ldquoTimefrequency parameters of the surface myoelectric signal forassessing muscle fatigue during cyclic dynamic contractionsrdquoIEEE Transactions on Biomedical Engineering vol 48 no 7 pp745ndash753 2001
[18] M McCaffery and A Beebe Pain Clinical Manual for NursingPractice V V Mosby Company Baltimore Md USA 1993
[19] M Gonzalez-Izal A Malanda I Navarro-Amezqueta et alldquoEMG spectral indices and muscle power fatigue duringdynamic contractionsrdquo Journal of Electromyography and Kine-siology vol 20 no 2 pp 233ndash240 2010
[20] G T Allison and T Fujiwara ldquoThe relationship between EMGmedian frequency and low frequency band amplitude changesat different levels of muscle capacityrdquoClinical Biomechanics vol17 no 6 pp 464ndash469 2002
[21] G R Naik D K Kumar V Yadav K Wheeler and S ArjunanldquoTesting of motor unit synchronization model for localizedmuscle fatiguerdquo in Proceedings of the 31st Annual InternationalConference of the IEEE Engineering in Medicine and BiologySociety (EMBS 09) pp 360ndash363 Minneapolis Minn USASeptember 2009
[22] A Cichocki and S-I Amari Adaptive Blind Signal and ImageProcessing John Wiley amp Sons New York NY USA 2002
[23] C D Meyer Matrix Analysis and Applied Linear AlgebraCambridge University Press Cambridge UK 2000
[24] M A Oskoei and H Hu ldquoSupport vector machine-basedclassification scheme for myoelectric control applied to upperlimbrdquo IEEE Transactions on Biomedical Engineering vol 55 no8 pp 1956ndash1965 2008
Figure 3 Average (plusmn SD) of six features of sEMG and level of contraction for the rested muscle (a) MDF (b) RMS (c) normalised spectralindex (NSM5) (d) IIS index (e) normalised RMS and (f) WL
Such an analysis is important for direct interpretation ofsEMG to estimate the force of muscle contraction
This study has experimentally studied and compared therelationship of muscle force and muscle fatigue with each ofthe six well-accepted features of sEMG Analysis of variance(ANOVA) was conducted to determine the effect of threemain different factors level of muscle contraction ( MVC)muscle fatigue and intersubject variation ANOVA resultsshow that RMS (119875 lt 005) and NSM5 (119875 lt 001) of sEMGwere significantly affected by the level of muscle contraction( MVC) The results also indicate that IIS index (119875 lt 001)was significant in identifying the muscle endurance limit
It is observed from the results that the intersubjectvariation was significant for RMS NRMS andWL while thevariation was not significant for NSM5 and IIS index (119875 gt005) Features such as RMS NRMS and WL are associatedwith the amplitude of the signal and the results indicate thatthere is a significant variation in the amplitude of sEMGbetween subjects However other features that are based onspectrum such as NSM5 or entropy dependent such as IISindex do not have significant differences between subjects
The linearity analysis indicates that the most linearrelationship of force of contraction (ranging from 50 to 100
MVC) was with NSM5 (1198772 = 096) followed by normalizedRMS (1198772 = 093) The linearity relationship between RMSof the signal and force of contraction was poor (1198772 = 071)This was also observed from the plots (Figure 2) The resultssuggest that the relationship between NSM5 a normalizedmeasure of the spectrumof the signal [12] and force ofmusclecontraction ( MVC) is the most significant and linear Theresults also indicate that for biceps NSM5 do not requireany normalization or calibrationThis indicates that NSM5 issuitable to estimate the level of muscle contraction comparedwith other features [6 24] even though this has not beenreported in literature
While earlier studies have identified NSM5 to be a fatigueindex [12 13] this study has shown that NSM5 of sEMGis the measure of force of muscle contraction This maybe attributed to the spectrum of sEMG being significantlyinfluenced by the rate of muscle stimulation and thus withforce of muscle contraction This study has also confirmedthat IIS index [14] is the most suitable indicator of musclebeing at the limit of endurance and is fatiguedThe study alsofound that the advantage of bothNSM5 and IIS was that thesedid not require any normalization
6 BioMed Research International
Conflict of Interests
The authors declare that there is no conflict of interestsregarding the publication of this paper
References
[1] J V Basmajian and C J de LucaMuscles Alive Their FunctionsRevealed by Electromyography Williams and Wilkins Balti-more Md USA 1985
[2] R Merletti and S Roy ldquoMyoelectric and mechanical manifes-tations of muscle fatigue in voluntary contractionsrdquo Journal ofOrthopaedic and Sports PhysicalTherapy vol 24 no 6 pp 342ndash353 1996
[3] T J Dartnall M A Nordstrom and J G Semmler ldquoMotor unitsynchronization is increased in biceps brachii after exercise-induced damage to elbow flexormusclesrdquo Journal of Neurophys-iology vol 99 no 2 pp 1008ndash1019 2008
[4] F E Marino M Gard and E J Drinkwater ldquoThe limits toexercise performance and the future of fatigue researchrdquo BritishJournal of Sports Medicine vol 45 no 1 pp 65ndash67 2011
[5] F A BainbridgeThePhysiology ofMuscular Exercise LongmanGreen and Co New York NY USA 3rd edition 1931
[6] S G Boe C L Rice and T J Doherty ldquoEstimating contractionlevel using root mean square amplitude in control subjects andpatients with neuromuscular disordersrdquo Archives of PhysicalMedicine and Rehabilitation vol 89 no 4 pp 711ndash718 2008
[7] RMerletti and P Parker Electromyography JohnWiley amp SonsNew York NY USA 2004
[8] D K Kumar N D Pah and A Bradley ldquoWavelet analysis ofsurface electromyography to determine muscle fatiguerdquo IEEETransactions on Neural Systems and Rehabilitation Engineeringvol 11 no 4 pp 400ndash406 2003
[9] D Moshou I Hostens G Papaioannou and H RamonldquoDynamicmuscle fatigue detection using self-organizingmapsrdquoApplied Soft Computing Journal vol 5 no 4 pp 391ndash398 2005
[10] S P Arjunan and D K Kumar ldquoDecoding subtle forearmflexions using fractal features of surface electromyogram fromsingle and multiple sensorsrdquo Journal of NeuroEngineering andRehabilitation vol 7 no 1 article 53 2010
[11] G Wang X-M Ren L Li and Z-Z Wang ldquoMultifractalanalysis of surface EMG signals for assessing muscle fatigueduring static contractionsrdquo Journal of Zhejiang University AScience vol 8 no 6 pp 910ndash915 2007
[12] G V Dimitrov T I Arabadzhiev K N Mileva J L Bowtell NCrichton andNADimitrova ldquoMuscle fatigue during dynamiccontractions assessed by new spectral indicesrdquo Medicine andScience in Sports and Exercise vol 38 no 11 pp 1971ndash1979 2006
[13] N A Dimitrova T I Arabadzhiev J-Y Hogrel and G V Dim-itrov ldquoFatigue analysis of interference EMG signals obtainedfrom biceps brachii during isometric voluntary contraction atvarious force levelsrdquo Journal of Electromyography and Kinesiol-ogy vol 19 no 2 pp 252ndash258 2009
[14] D K Kumar S P Arjunan andG R Naik ldquoMeasuring increasein synchronization to identify muscle endurance limitrdquo IEEETransactions on Neural Systems and Rehabilitation Engineeringvol 19 no 5 pp 578ndash587 2011
[15] S Minning C A Eliot T L Uhl and T R Malone ldquoEMGanalysis of shoulder muscle fatigue during resisted isometricshoulder elevationrdquo Journal of Electromyography and Kinesiol-ogy vol 17 no 2 pp 153ndash159 2007
[16] D Farina A Holobar R Merletti and R M Enoka ldquoDecodingthe neural drive to muscles from the surface electromyogramrdquoClinical Neurophysiology vol 121 no 10 pp 1616ndash1623 2010
[17] P Bonato S H Roy M Knaflitz and C J de Luca ldquoTimefrequency parameters of the surface myoelectric signal forassessing muscle fatigue during cyclic dynamic contractionsrdquoIEEE Transactions on Biomedical Engineering vol 48 no 7 pp745ndash753 2001
[18] M McCaffery and A Beebe Pain Clinical Manual for NursingPractice V V Mosby Company Baltimore Md USA 1993
[19] M Gonzalez-Izal A Malanda I Navarro-Amezqueta et alldquoEMG spectral indices and muscle power fatigue duringdynamic contractionsrdquo Journal of Electromyography and Kine-siology vol 20 no 2 pp 233ndash240 2010
[20] G T Allison and T Fujiwara ldquoThe relationship between EMGmedian frequency and low frequency band amplitude changesat different levels of muscle capacityrdquoClinical Biomechanics vol17 no 6 pp 464ndash469 2002
[21] G R Naik D K Kumar V Yadav K Wheeler and S ArjunanldquoTesting of motor unit synchronization model for localizedmuscle fatiguerdquo in Proceedings of the 31st Annual InternationalConference of the IEEE Engineering in Medicine and BiologySociety (EMBS 09) pp 360ndash363 Minneapolis Minn USASeptember 2009
[22] A Cichocki and S-I Amari Adaptive Blind Signal and ImageProcessing John Wiley amp Sons New York NY USA 2002
[23] C D Meyer Matrix Analysis and Applied Linear AlgebraCambridge University Press Cambridge UK 2000
[24] M A Oskoei and H Hu ldquoSupport vector machine-basedclassification scheme for myoelectric control applied to upperlimbrdquo IEEE Transactions on Biomedical Engineering vol 55 no8 pp 1956ndash1965 2008
The authors declare that there is no conflict of interestsregarding the publication of this paper
References
[1] J V Basmajian and C J de LucaMuscles Alive Their FunctionsRevealed by Electromyography Williams and Wilkins Balti-more Md USA 1985
[2] R Merletti and S Roy ldquoMyoelectric and mechanical manifes-tations of muscle fatigue in voluntary contractionsrdquo Journal ofOrthopaedic and Sports PhysicalTherapy vol 24 no 6 pp 342ndash353 1996
[3] T J Dartnall M A Nordstrom and J G Semmler ldquoMotor unitsynchronization is increased in biceps brachii after exercise-induced damage to elbow flexormusclesrdquo Journal of Neurophys-iology vol 99 no 2 pp 1008ndash1019 2008
[4] F E Marino M Gard and E J Drinkwater ldquoThe limits toexercise performance and the future of fatigue researchrdquo BritishJournal of Sports Medicine vol 45 no 1 pp 65ndash67 2011
[5] F A BainbridgeThePhysiology ofMuscular Exercise LongmanGreen and Co New York NY USA 3rd edition 1931
[6] S G Boe C L Rice and T J Doherty ldquoEstimating contractionlevel using root mean square amplitude in control subjects andpatients with neuromuscular disordersrdquo Archives of PhysicalMedicine and Rehabilitation vol 89 no 4 pp 711ndash718 2008
[7] RMerletti and P Parker Electromyography JohnWiley amp SonsNew York NY USA 2004
[8] D K Kumar N D Pah and A Bradley ldquoWavelet analysis ofsurface electromyography to determine muscle fatiguerdquo IEEETransactions on Neural Systems and Rehabilitation Engineeringvol 11 no 4 pp 400ndash406 2003
[9] D Moshou I Hostens G Papaioannou and H RamonldquoDynamicmuscle fatigue detection using self-organizingmapsrdquoApplied Soft Computing Journal vol 5 no 4 pp 391ndash398 2005
[10] S P Arjunan and D K Kumar ldquoDecoding subtle forearmflexions using fractal features of surface electromyogram fromsingle and multiple sensorsrdquo Journal of NeuroEngineering andRehabilitation vol 7 no 1 article 53 2010
[11] G Wang X-M Ren L Li and Z-Z Wang ldquoMultifractalanalysis of surface EMG signals for assessing muscle fatigueduring static contractionsrdquo Journal of Zhejiang University AScience vol 8 no 6 pp 910ndash915 2007
[12] G V Dimitrov T I Arabadzhiev K N Mileva J L Bowtell NCrichton andNADimitrova ldquoMuscle fatigue during dynamiccontractions assessed by new spectral indicesrdquo Medicine andScience in Sports and Exercise vol 38 no 11 pp 1971ndash1979 2006
[13] N A Dimitrova T I Arabadzhiev J-Y Hogrel and G V Dim-itrov ldquoFatigue analysis of interference EMG signals obtainedfrom biceps brachii during isometric voluntary contraction atvarious force levelsrdquo Journal of Electromyography and Kinesiol-ogy vol 19 no 2 pp 252ndash258 2009
[14] D K Kumar S P Arjunan andG R Naik ldquoMeasuring increasein synchronization to identify muscle endurance limitrdquo IEEETransactions on Neural Systems and Rehabilitation Engineeringvol 19 no 5 pp 578ndash587 2011
[15] S Minning C A Eliot T L Uhl and T R Malone ldquoEMGanalysis of shoulder muscle fatigue during resisted isometricshoulder elevationrdquo Journal of Electromyography and Kinesiol-ogy vol 17 no 2 pp 153ndash159 2007
[16] D Farina A Holobar R Merletti and R M Enoka ldquoDecodingthe neural drive to muscles from the surface electromyogramrdquoClinical Neurophysiology vol 121 no 10 pp 1616ndash1623 2010
[17] P Bonato S H Roy M Knaflitz and C J de Luca ldquoTimefrequency parameters of the surface myoelectric signal forassessing muscle fatigue during cyclic dynamic contractionsrdquoIEEE Transactions on Biomedical Engineering vol 48 no 7 pp745ndash753 2001
[18] M McCaffery and A Beebe Pain Clinical Manual for NursingPractice V V Mosby Company Baltimore Md USA 1993
[19] M Gonzalez-Izal A Malanda I Navarro-Amezqueta et alldquoEMG spectral indices and muscle power fatigue duringdynamic contractionsrdquo Journal of Electromyography and Kine-siology vol 20 no 2 pp 233ndash240 2010
[20] G T Allison and T Fujiwara ldquoThe relationship between EMGmedian frequency and low frequency band amplitude changesat different levels of muscle capacityrdquoClinical Biomechanics vol17 no 6 pp 464ndash469 2002
[21] G R Naik D K Kumar V Yadav K Wheeler and S ArjunanldquoTesting of motor unit synchronization model for localizedmuscle fatiguerdquo in Proceedings of the 31st Annual InternationalConference of the IEEE Engineering in Medicine and BiologySociety (EMBS 09) pp 360ndash363 Minneapolis Minn USASeptember 2009
[22] A Cichocki and S-I Amari Adaptive Blind Signal and ImageProcessing John Wiley amp Sons New York NY USA 2002
[23] C D Meyer Matrix Analysis and Applied Linear AlgebraCambridge University Press Cambridge UK 2000
[24] M A Oskoei and H Hu ldquoSupport vector machine-basedclassification scheme for myoelectric control applied to upperlimbrdquo IEEE Transactions on Biomedical Engineering vol 55 no8 pp 1956ndash1965 2008
