EMG signal amplitude normalization technique in stretch-shortening cycle movements

Joumal of Eiecfmn ography md Kinesioiogy Vol. 3, No. 4, pp 2Y6-244 $l~~lhtt~utt;t&ememann Ltd

EMG Signal Amplitude Normalization Technique in Stretch-shortening Cycle Movements G. T. Allisod,

R. N. MarshalP,

and K. P. Singer’

‘School of Physiotherapy, Curtin University, Shenton Park, W. Australia 6008 and 2Department of Human Movement, The University of Western Australia, Nedlands, W. Australia 6009, Australia

Summary: Analysis of functional movements using surface electromyography (EMG) often involves recording both eccentric and concentric muscle activity during a stretch-shorten cycle (SSC). The techniques used for amplitude normalization are varied and are independent of the type of muscle activity involved. The purpose of this study was: (i) to determine the effect of 11 amplitude normalization techniques on the coefficient of variation (CV) during the eccentric and concentric phases of the SSC; and (ii) to establish the effect of the normalization techniques on the EMG signal under variable load and velocity. The EMG signal of the biceps brachii of eight normal subjects was recorded under four SSC conditions and three levels of isometric contraction. The 11 derived normalization values were total rms, mean rms and peak rms (100 ms time constant) for the isometric contractions and the mean rms and peak rms values of the ensemble values for each set of isotonic contractions. Normalization using maximal voluntary isometric contractions (MVIC), irrespective of rms processing (total, mean or peak), demonstrated greater CV above the raw data for both muscle actions. Mean ensemble values and submaximal isometric recordings reduced the CV of concentric data. NO amplitude normalization technique reduced the CV for eccentric data under loaded conditions. An ANOVA demonstrated significant (PcO.01) main effects for load and velocity on concentric raw data and an interaction (P
INTRODUCTION

The analysis of functional movements often involves recording both eccentric and concentric muscle activity during a stretch-shortening cycle (SSC). Surface electromyography (EMG) has been extensively used to analyse the SSP5. Mean

Received April 21, 1993. Revised July 14, 1993. Accepted September 8, 1993. Address correspondence and reprint requests to Mr G. T. Allison, School of Physiotherapy, Cm-tin University, Selby Street, Shenton Park, W. Australia 6008, Australia.

236

237

EMG AMPLITUDE NORMALIZATION electrical activity of specific muscle groups during functional tasks has been calculated over multiple trials to establish normal profile@,‘. These descriptive ‘templates’ are often compared with those of individuals with injury or pathology to identify abnormalities in muscle activity. For example, in pathological SSCs of gait, abnormalities in both phase and amplitude characteristics of the EMG signal are evident1,8,9. To assist in developing EMG templates, amplitude normalization techniques have been used to reduce the inter-individual variability of grouped data. This process however assumes that the normalization transformation is independent of the type of muscle action during a SSC. Therefore, it is important to ensure that any amplitude normalization technique maintains the true biological variance, occurring with either concentric or eccentric muscle actions, while reducing the individual variation inherent in surface EMG signal. Amplitude Normalization Procedures Amplitude normalization procedures are used when recording surface EMG signals because there is inherent physiological variability between individuals. Secondly variability associated with electrode replacement may alter the electrical impedance and the spatio-temporal relationships between the active motor units and the recording electrode. Individual subject data are commonly amplitude normalized before calculating group means, and as a result the mean and variance of the original data may be transformed. For this reason, the coefficient of variation (CV), which expresses the ratio of the standard deviation (SD) as a percentage of the mean, is often used to compare transformed data. The effectiveness of an amplitude normalization technique has been based on its ability to reduce the CV of the processed data when compared to the non-normalized or ‘raw’ datalo. However, amplitude normalized data from normal gait demonstrates large CV ranging from 31 to 198% depending on the muscle being recorded and the normalization technique employed l”*ll. Choosing the reduction in the variance as the sole selection criterion for a normalization technique is not appropriate if the technique removes true biological variance manifest between (and within) individualsl. Similarly, using particular techniques which increase the variation of the grouped data may also be inappropriate. In the literature, processed surface EMG signals

from a maximal voluntary isometric contraction (MVIC) are commonly used for amplitude normalizationl*‘. It has been argued that using the EMG signal from a MVIC for amplitude normalization provides data that expresses the relative contractile activity of the individual’s muscle during the complete task9,10. Yet, it has been shown that the EMG signal of the MVIC is less reliable than submaximal isometric contractions; using the MVIC to amplitude normalize data may increase the CV. Further, maximal EMG signal amplitude may not reflect equivalent levels of force or activation during eccentric or concentric phases of the SSC4,5,12.‘3. For analysis of gait, amplitude normalization with the ensemble peak and/or mean EMG signal of multiple or single trials has demonstrated a reduced CV of the grouped data lJ”. On this basis, these techniques have been used to reduce intra- and inter-individual variation in surface EMG data for dynamic tasks 6,7,9Jo. Nevertheless, there remains significant diversity among the amplitude normalization approaches currently reported in the literature. Eccentric and Concentric Muscle Actions Most functional task analyses by electromyography involve the SSC and often incorporate specific focus on the eccentric and concentric muscle actions2-5,14. Each type of muscle activity has unique neurophysiological and mechanical characteristics. For example, the amplitude of the resultant EMG signal at constant tensions has been shown to be greater during the concentric phase when compared to the eccentric phase4,5J4. Furthermore, there is a strong relationship between the EMG signal amplitude and the load placed on the muscle, yet these relationships differ between eccentric and concentric phases of the SSC under increasing load and velocity 5~13,15 . This is particularly true for the mechanical efficiency as load increases4. The specific analysis of the eccentric or concentric phase of a SSC is not reflected in specific techniques of amplitude normalization. This implies that amplitude normalization techniques are independent of the type of muscle action, velocity and load. Such an assumption would seem tenuous in the light of the above differences, and has yet to be addressed in the literature. AIMS The first aim of this study was to determine the effect of amplitude normalization techniques on the Journal of Electromyography

& Kinesiology Vol. 3, No. 4, 1993

238

G. T. ALLISON

overall inter-individual variation during the eccentric and concentric phases of the SSC. The second aim was to establish the effect of the normalization techniques on the amplitude of the EMG signal of the biceps brachii under variable load and velocity. METHODS Subjects Eight normal subjects (three male, five female) provided informed consent prior to testing then attended two testing sessions on consecutive days. On the first day a general familiarization session and then a practice of the experimental procedure was undertaken. On the second day the procedures were repeated and the data obtained used for further analysis.

ET AL.

practised using a metronome to maintain a steady frequency during each sequence. Electrode Configuration The biceps and triceps brachii and the acromion process were prepared on the dominant arm of the subject with maximal electrode impedance less than 3 K ohms. The anterior aspect of the acromion process was prepared as a common earth. The long head of biceps brachii and the lateral head of the triceps brachii had three 3M Ag/AgCl electrodes positioned at an inter-electrode distance of 20 mm aligned with the respective muscle bellies. The two outside electrodes were shorted in a method described by Koh and Grabiner16 to minimize crosstalk artifact. Signal Analysis

Test Protocol Each subject was positioned on a stable, adjustable chair with their feet clear of the floor and arm and trunk supported from the axilla to the elbow. During the isometric contractions the forearm, in mid supination-pronation, was horizontal with an elbow angle of 60”. After a brief familiarization session each subject was required to perform isometric elbow flexion contractions at three intensities and elbow flexion-extension cycles under four loaded conditions. The isometric contractions preceded the SSC in all subjects yet the order within each set were randomized. Five trials for each isometric contraction were performed. There were two submaximal contractions: (i) Medical Research Councill’, Grade III; and (ii) a 2.3 kg load held in the hand. Each subject practised the submaximal isometric contractions with special attention to minimizing the triceps EMG signal. The triceps brachii EMG signal trace was displayed on an oscilloscope as visual feedback. Each trace was compared with each subject’s resting activity of the triceps to ensure minimal co-contraction during the submaximal isometric contractions. The other isometric contraction was maximal with a gradual build-up of force against a resistance from a 5 cm wide nylon tether, adjusted for each individual at the distal radius. The isotonic movements were full elbow flexion-extension with and without a 2.3 kg load held in the hand, completed at two frequencies: 20 and 40 cycles per min respectively. The activity was Journal

ofElecwomyography

& Kinesiology Vol. 3, No. 4, 1993

Surface EMG data from the biceps brachii was high- and low-pass filtered (8 Hz and 800 Hz), amplified and recorded at 1024 Hz over five trials. The biceps brachii EMG data were stored on disc using a Macintosh II PC, equipped with a 32-bit AD board, running Superscope@ software. One second of data was recorded and stored on disc for each of the five isometric contractions. During the isotonic contractions, EMG signal data were recorded for a minimum of five continuous cycles for both the fast and slow movement frequencies. All the amplified raw data were then demeaned and full wave rectified (FWR). Isometric amplitude normalization values were then calculated from the isometric FWR data. This included calculating the total root mean square (rms) of the EMG signal and the mean and peak values of the rms linear envelope created over a 100 ms time constant. The isotonic data were then processed using a 100 ms time constant rms processing technique to create a linear envelope (LE) from which the ensemble amplitude normalization data (mean and peak values) were derived. Each SSC was defined by an electronic event marker recorded simultaneously and was then time normalized to 100% of the full extension-flexion cycle. The time normalization program compressed the LE data to a set number of data points (0.5% or 200 points) by calculating the mean for that percentage of the total time of the trial. The mean of five cycles was calculated for each subject for each of the four isotonic contraction protocols.

EMG AMPLITUDE The subjects’ means were then combined to establish the group means and standard deviations. From this data CV were calculated for the whole group for each 0.5% of the total cycle. This was defined as the non-normalized or ‘raw data’ and represented the group’s variation. Each subject’s trial means were then amplitude normalized using nine isometric amplitude normalization values and two isotonic ensemble amplitude normalization values. Eccentric and concentric phases were defined as the set of 50 data points (quartiles) of the time normalized data starting at 12.5% and 62.5% respectively, in order to eliminate any problems associated with movement end points. The data were compared (t test) for differences between the means of the EMG signal amplitudes for the eccentric and concentric phases. However, since each concentric and eccentric phase was only part of the total SSC, it was assumed that these data sets were not independent and did not necessarily reflect matched muscle length-tension relationships. Consequently, further analysis of the concentric and eccentric data was made separately. The effect of velocity and load on the amplitude of the EMG data during both eccentric and concentric contractions was made using a 2 (no-load and loaded) x 2 (fast and slow) repeated measures analysis of variance. For all statistical tests a probability level of PcO.05 was adopted as the criterion of significant differences. RESULTS Raw Data The concentric and eccentric quartiles of the time normalized EMG data of the biceps brachii during four stretch-shortening cycle conditions are illustrated in Figure 1. The EMG signal amplitude means and standard deviations of the concentric and eccentric phases are recorded in Tables 1 and 2 respectively. These results demonstrated an increased amplitude during the concentric SSC phase when compared to the subsequent eccentric phase. Since the phases are not independent further analysis was made within each phase separately. Tables 3 and 4 show the F ratios generated from a 2 (no-load and loaded) x 2 (fast and slow) ANOVA for both forms of muscle activity. The raw concentric data demonstrated significant increases under both increased velocity and intensity

239

NORMALIZATION

(Table 3). The eccentric raw data however demonstrated a significant interaction between the ioad and velocity (Table 4).

Amplitude Normalized Data A summary of the repeated measures analysis of variance (F ratios) results for the effects of load and velocity on the processed amplitude of the EMG signal are shown in Tables 3 (concentric data) and 4 (eccentric data). These values demonstrated that data transformation due to amplitude normahzation may alter ANOVA findings. Mean rms processing techniques altered the F ratio in the concentric data such that velocity was not a significant factor, i.e., the statistical analyses lost the sensitivity to the effects of changes in velocity. Furthermore, for concentric data, there were statistical interactions between velocity and load after peak rms normalization was performed (Table 3).

Coefficients of Variation Figure 2 compares the CV expressed as a decimal of all 11 amplitude normalization values for both eccentric and concentric phases for four SSC conditions. The CV of the raw data is greatest during the slow no-load tasks and similar in the other three conditions. Similar CVs for the eccentric and concentric phases occur in all four SSC conditions with eccentric showing greater standard deviation as a ratio of the mean amplitude during the faster movements. Comparisons between the raw CV and the corresponding CV values for the amplitude normalization values are also illustrated in Figure 2. It would seem that the EMG processing technique (i.e. total rms, mean and peak rms) have little influence on the overall CV and it is the relative intensities of the contraction which demonstrate the greatest influence. When compared with the raw data isometric grade III has little effect in reducing the CV. The 2.3 kg load demonstrated decreases in the CV of the concentric phase but increases the eccentric CV in all cases except the slow no-load condition. Amplitude normalization using maximal isometric contractions increases the CV as compared to the raw data in almost all conditions. Journalof Electromyography

& Kinesiology

Vol. 3, No. 4, 1993

G. T.ALLISON ETAL.

240

0.4concentric Q

0.3-

Slow

D

eccentric

4

a

D

O.l0

Fast

37.5%

12.5%

62.5%

Time Normal&d

87.5%

(9’0)

FIG. 1. Eccentric and concentric quartiles of a, slow load, slow no-load; b, fast load and fast no-load from EMG linear envelope. Signal was time normalized to 200 data points from 100 ms rms linear envelope. _ Loaded; unloaded.

TABLE 1.

Concentric

Normalization process Raw Total rms Grade Ill 2.3 kg load MVIC Mean rms Grade Ill 2.3 kg load MVIC Peak rms Grade Ill 2.3 kg load MVIC Ensemble rms Peak Mean

Journal

of Electromyography

EMG signal

Mean

means and standard deviations for raw and amplitude under four testing protocols Testing

Slow

No-load

Load

protocol

(SD)

Mean

(so)

Mean

0.038

0.042

0.157

0.093

2.461 0.221 0.061

2.551 0.115 0.070

9.700 1.079 0.269

2.467 0.215 0.061

2.643 0.125 0.071

9.451 1.044 0.252

1.773 0.166 0.047

1.751 0.079 0.052

0.085 0.085

0.049 0.049

& Kinesiology Vol. 3, No. 4, 1993

normalized data

Fast

No-load

Load

(so)

Mean

0.040

0.019

0.200

0.108

6.143 0.561 0.234

2.775 0.281 0.069

1.909 0.109 0.051

13.154 1.377 0.333

8.619 0.654 0.243

6.093

0.540 0.210

2.820 0.285 0.070

1.913 0.110 0.051

9.231 0.987 0.273

6.178 0.480 0.170

7.221 0.828 0.219

4.701 0.484 0.210

1.954 0.201 0.050

1.376 0.080 0.037

13.380 1.399 0.338

8.644 0.666 0.247

0.426 2.297

0.275 0.701

0.106 0.513

0.050 0.092

0.521 1.076

0.329 0.346

(so)

EMG AMPLITUDE TABLE 2.

241

NORMALIZATION

Eccentric EMG signal means and standard deviations for raw and amplitude under four testing protocols Testing

normalized

data

protocol Fast

Slow

Load

No-load

Load

No-load

Normalization process

Mean

(so)

Mean

(so)

Mean

(so)

Mean

(so)

Raw

0.021

0.021

0.058

0.027

0.020

0.011

0.082

0.044

Total rms Grade Ill 2.3 kg load MVIC

1.289 0.125 0.033

1.239 0.088 0.035

3.871 0.456 0.092

2.131 0.353 0.058

1.155 0.155 0.032

0.622 0.121 0.027

6.256 0.688 0.129

4.980 0.616 0.073

1.295 0.122 0.033

1.287 0.093 0.036

3.828 0.449 0.087

2.216 0.363 0.053

1.177 0.158 0.033

0.634 0.123 0.027

4.371 0.475 0.106

3.518 0.384 0.050

0.930 0.093 0.025

0.853 0.061 0.026

2.814 0.331 0.074

1.478 0.225 0.053

0.809 0.109 0.023

0.437 0.080 0.019

6.361 0.699 0.131

5.023 0.626 0.074

0.049 0.049

0.032 0.032

0.167 0.438

0.101 0.258

0.059 0.269

0.035 0.142

0.236 2.245

0.169 0.715

Mean rms Grade III 2.3 kg load MVIC Peak rms Grade III 2.3 kg load MVIC Ensemble Peak Mean

rms

TABLE 3. F ratios for raw and amplitude normalized concentric data comparing EMG signal amplitude for two load ho-load and load) and two velocities (fast and slow) testing conditions Normalization value

Load F

Velocity F

20.979b

15.404b

Load*

TABLE 4. F ratios for raw and amplitude normalized eccentric data comparing EMG signal amplitude for two load (no-load and load) and two velocities (fast and slow) testing conditions Normalization value

Velocity F

Load F

Velocity F

Load* Velocity F

2.278

Raw

19.636b

4.001

6.472”

5.104 12.922b 13.941b

3.337 6.434” 1.732

Total rms Grade Ill 2.3 kg MVIC

10.525b 8.286” 19.064’=

3.221 3.884 5.287

4.259 5.823” 6.371”

15.254’= 22.649b 13.307b

0.015 0.043 0.996

0.147 2.377 0.051

Mean rms Grade Ill 2.3 kg MVIC

9.302” 8.448” 22.098b

0.255 1.071 1.202

0.633 0.040 1.310

Peak rms Grade Ill 2.3 kg MVIC

18.714b 28.314’= 10.874”

8.991” 17.719b 11.328”

9.217” 15.108” 6.760”

Peak rms Grade Ill 2.3 kg MVIC

11.599” 9.256” 19.339b

5.885” 4.962 11.838”

7.021” 6.598” 13.735b

Ensemble Mean Peak

70.65gb 1 5.678b

13.458” 7.880”

2.660 3.268”

Ensemble Mean Peak

58.704b 12.156”

39.383’= 4.492

49.127b 7.514”

Raw Total rms Grade Ill 2.3 kg MVIC

17.128” 26.377” 11.043b

Mean rms Grade Ill 2.3 kg MVIC

“RO.05; bP
(cell means

and standard

deviations

in

DISCUSSION Raw Data

The findings of this study support the wellestablished relationship that under equivalent tensions the amplitude of the EMG signal is greater

“RO.05; “RO.01 Table 2).

(cell means and standard

deviations

in

during concentric muscle action when compared with eccentric activity5. Furthermore, as load increases the concentric phase increases are proportionally greater4. It is of interest to note that the effect of velocity is significant although the data were time Journal of Electromyography

& Kinesiology Vol. 3. No. 4, 1993

242

G. T. ALLISON Raw Data

Isometric Grade III

Isometric 2.3kg load

l.PJ--

ET AL. MVIC

Isotonic Ensemble 1-d

1 0.8 0.6 0.4 0.2

FIG. 2. Mean coefficient of variations (CV) for concentric and eccentric data for raw and 11 amplitude normalization techniques for four SSC conditions. EMG signal amplitude values used to normalize were total, mean and peak rms calculated over 100 ms for each of three isometric force levels, with mean and peak rms calculated for each of the SSC conditions. 0 Eccentric activity; n concentric activity.

normalized. This is different to the reported effect time normalization has on integrated EMG during a fatiguing SSC task of the lower limb3. This may be explained by the fact that this study analysed quartiles of a non-fatigued SSC using a rms processing technique. The raw data demonstrates that the amplitude of the EMG signal reflects the activation of a muscle relative to the resistance (load and velocity) placed on that muscle. However this relationship is different for the concentric and eccentric phases of the SSC. The significant increase of the concentric EMG JournalofEiectromyography

& Kinesiology Vol. 3, No. 4, 1993

signal amplitude due to the 2.3 kg load and increased velocity is well established13,15. However in this study the raw eccentric data demonstrated an interaction between velocity and load. This can be interpreted in that the difference between the EMG signal amplitude of the loaded and unloaded contractions is significantly greater during the faster movement when compared to the difference during the slow movement, The reasons for the differences between the findings from the raw concentric and the raw eccentric data are not clear. They may be explained by the influence of the series elastic

EMGAMPLiTULlE component, mechanical efficiency and neural influences during contractions of variable loads and velocities2-5. Nevertheless, since there is a significant interaction, further generalizations as to the variance contributed by load or velocity independently are tenuous. Amplitude Normalized Data The amplitude normalized data in both eccentric and concentric phases does not show the same relationships as demonstrated in the raw EMG data. Amplitude normalization using total rms of a grade III isometric contraction and mean rms values for all three isometric contractions (grade III, submaximal and MVIC) demonstrate no statistically significant difference between the EMG signal amplitude of the fast and slow SSC. In this study, the amplitude normalized data fails to demonstrate changes in muscle activation as reflected by EMG signal amplitude changes due to an increase in velocity. This shows a loss of sensitivity that may lead to possible errors in analysis of the interaction of EMG signal amplitude and velocity of contraction during the SSC. An analogy of this in the experimental setting is the comparison of EMG amplitude templates during variable gait cadence. Clearly, these normalization procedures have either removed true biological variance or reduced the power of the experimental procedure to produce a type II error.

243

NORMALIZATION

of the amplitude normalization technique utilized in this study. It would seem therefore that under load or increased velocity, amplitude normalization techniques increase the CV of EMG data during the eccentric phase of the SSC. These results may explain the larger CV exhibited by two-joint muscles (biceps femoris and rectus femoris) during EMG signal analysis of gait when compared to one-joint muscles’O. However, it has yet to be shown if this is due to the type of contraction, muscle length characteristics and/or a component of the amplitude normalization process. Possible mechanisms for the differences between eccentric and concentric data are the series elastic components, changes in mechanical efficiency and neural influences. These factors may be significant in the function of two-joint muscles and eccentric muscle actions and may account for significant variance between individuals which are independent of isometric or ensemble normalization values. The increased variation due to the series elastic component and changes in mechanical efficiency are further supported by the fact that the CV during the eccentric phase is reduced only under the slow no-load testing condition. Under these conditions both the series elastic component and changes in mechanical efficiency are minimized for the four conditions tested in this study. This is confirmed by the similarity of the CV values for both concentric and eccentric muscle actions during the slow noload tasks when compared with the other conditions tested.

Coefficient of Variation Analysis Our findings demonstrate that, for the biceps brachii under SSC testing conditions, MVIC amplitude normalization processing increases the CV of the data above that of the raw data. This would seem to be unrelated to the type of rms processing involved since similar values were generated irrespective of the processing technique (total, mean or peak rms values). In contrast, it is clear that the mean ensemble and the submaximal set load amplitude normalization processes reduce the CV during the concentric phase of the SSC under all conditions tested. Similar reductions of the CV using the same normalization values occurred in the eccentric component of the SSC only during the slow unloaded tasks. In the other three testing formats (fast load, slow load and fast no-load) eccentric CV were generally increased irrespective

CONCLUSIONS It is necessary to establish the purpose of amplitude normalization procedures since in ‘normalizing data’ often there is a concomitant loss of information. Therefore the use of reduced CV as the sole selection criteria for normalization techniques must be treated with caution, since improved. statistical power may be at the expense of true biological variance. In this experiment, this would seem to be the case for velocity information when EMG signal data is amplitude normalized using mean rms values for isometric contractions of any intensity or total rms for grade III isometric contractions. The use of MVIC increases the CV within a group yet maintains the fundamental characteristics between load, velocity and the EMG signal amplitude for a SSC. In comparison, total or peak rms Journal

ofElectmmyography

& Kinesiology

Vol. 3. No. 4, 1993

244

G. T. ALLISON

values, using a set submaximal load as a source for amplitude normalization, effectively reduce the CV while maintaining the characteristics of the original data. Furthermore, previous research has demonstrated that the EMG signal from a submaximal load is more reliable than the signal from a MVIC and in the clinical setting may be safer than MVIC. Nevertheless, despite the findings of this study, MVIC may continue to be used for the benefit of relative intensity levels. Similarly, it would seem that on the basis of reduced CV and maintenance of the significant F ratios the mean ensemble value could be recommended for EMG signal amplitude normalization. This may be at_ the expense of important factors associated with inter-individual variation. For amplitude normalization techniques which use isometric contractions, a submaximal isometric load reduces the CV for all tasks during the concentric phase, however for the eccentric phase it is less clear since it is only true for slow unloaded conditions. Therefore, focused analysis of the eccentric phase of the SSC, as compared with the concentric phase, requires specific consideration of the amplitude normalization technique utilized. It is not clear what mechanism(s) affect this relationship, however the series elastic components, mechanical efficiency and levels of central activation may play significant roles.

ET AL. neuromuscular activation patterns of human skeletal muscle. 1987. Komi PV, Kaneko M, Aora 0: EMG activity of the leg extensor muscle with special reference to mechanical efficiency in concentric and eccentric exercise. Int J Sportr Med 8(suppl.):22-29, 1987. Komi PV: Stretch-shortening cycle. In: Strength and Power in Sport, Vol. 3, ed by Komi PV, Blackwell Scientific Publications. Melbourne. DD. 169-179. 1992. Kadaba MP: Ramakrishnan HK, Wootten ME, Gainey J, Gorton G, Cochran GVB: Repeatability of kinematic, kinetic, and electromyographic data in normal adult gait. J Orthopaed Res 7849-860, 1989. Dubo HIC, Peat M, Winter DA, Quanbury AO, Hobson DA, Steinke T, Reimer G: Electromyographic temporal analysis of gait: normal human locomotion. Arch Phys Med Rehab 57:415-420, 1976. Peat M, Dubo HIC, Winter DA, Quanbury AO, Steinke T, Grahame R: Electromyographic temporal analysis of gait: Hemiplegic locomotion. Arch Phys Med Rehab, 57:421425, 1976. Knutsson E, Richards C: Different types of disturbed motor control in gait of hemiparetic patients. Brain 102X&420, Znt J Sports Med 8(suppl.):3w7,

4.

5.

6.

7.

8.

9.

1979.

10. Yang JF, Winter DA: Electromyography amplitude normalisation methods: improving their sensitivity as diagnostic tools in gait analysis. Arch Phys Med Rehab 65:517-521, 1984. 11. Van Woensel W. Van Der Helm F. Van Der Hoeven M. Daanen H: Shoulder muscle EMG in normals as a reference for clinical evaluation. In: Electromyographical Kinesiology. Proceedings of the 8th Congress of International Society of Electrophysiological Kinesiology, ed by Anderson PA, Hobart DJ, Danoff JV, Baltimore MD, 12-16 Aug 1990. Amsterdam, Exerpta Medica, 1991. 12. Yang JF, Winter DA: Electromyography reliability in maximal and submaximal isometric contractions. Arch Phys Med Rehab 64:417-420, 1983. 13. White SC, Winter DA: Predicting muscle force in gait from EMG signals and musculotendon kinematics. J Electromyogr Kinesiol2:217-231,

REFERENCES 1. Arsenault AB, Winter DA, Martenuik RG: Characteristics of muscular function and adaptation in gait: a literature review. Physiorirerapy Curtada 395-13, 1987. 2. Gollhofer A, Komi PV, Miyashita M, Aura 0: Fatigue during stretch-shortening cycle exercises: Changes in mechanical performance of human skeletal muscle. Znr / Sports Med 871-78, 1987. 3. Gollhofer A, Komi PV, Fujitsuka N, Miyashita M: Fatigue during stretch-shortening cycle exercises. II. Changes in

Journal of Eiectromyography & Kinesiology Vol. 3, No. 4, 1993

1993.

14. Winter DA: Biomechanics and Motor Control of Human Movement, (2nd edn.), New York, John Wiley & Sons Inc, 1990. 15. Solomonow M, Baratta R, Shoji H, D’Ambrosia R. The EMG-force relationships of skeletal muscle; dependence on contraction rate, and motor unit control strategy. Electromyogr

Clin. Neurophysiol

30:141-152,

1990.

16. Koh TJ, Grabiner MD: Evaluation of methods to minimize electromyography. J Biomech crosstalk in surface 26:S151-S157,

1993.

17. Medical Research Council: Aids to the examination of the peripheral nervous system. Memorandum No. 45. London, HMSO, 1980.