Folia phoniat. 27: 177- 189 (1975)
A Study of the Electromyographic Activity of the Muscle Orbicularis Oris A. W. Kelman and S. Gatehouse West of Scotland Health Boards, Department of Clinical Physics and Bio-Engineering, Glasgow
Introduction The relevance of electromyographic (EMG) data to phonetics has already been recognised, and the EMG activity of the facial muscles, while the subject phonated a series of consonant-vowel-consonant (CVC) or vowel-consonantvowel (VCV) utterances has been investigated (1, 2, 6 -9 ). In one series of experiments, Leanderson el al. (4, 5) investigated several facial muscles simulta neously, using needle electrodes. However, most of the research has utilised surface electrodes placed on the muscle orbicularis oris (m. orb. oris). Since different electrode sites and configurations have been used, e.g. upper lip or lower lip, on midline or off midline, it is very difficult to correlate results obtained from different experiments. Accepting the fact that surface electrodes measure activity from the surrounding muscle and not from a single or small group of motor units, it seemed evident that a study of the EMG output asso ciated with different electrode positions was required. A series of experiments was thus undertaken in which the EMG activity could be measured simultaneously from four electrode sites in the configurations A, B and C (fig. 1). The electrodes could be connected in either differential (bipolar) pairs or in a monopolar mode, with the reference electrode situated on the mastoid. The subject was asked to repeat many times the utterances: Ip Ip/, /pAp/, /plb/, /bib/, /blp/, /bAb/. These were chosen so as to investigate whether there were any significant differences in the EMG activity produced by the m. orb. oris for the labial sounds / p/ and /b/, in different phonetic contexts.
Downloaded by: King's College London 137.73.144.138 - 3/6/2018 9:04:34 PM
Received: February 7, 1975; accepted: March 9, 1975.
Kelman/Gatehouse
178
In this paper we report on the first experiment in which the utterance /pip/ was used and where the objectives were much more limited, namely, using the electrode configurations shown in figure 1 A with the electrodes connected in differential mode in the sequence 1-2, 1-3 , 2 -4 , and 3 -4 . (1 ) Were there any significant differences between the EMG activity mea sured by electrode pairs 1 -2 and 1 -3 , and also between 2 - 4 and 3 -4 , i.e. was there left-right symmetry about the vertical midline? (2) Were there any significant differences between the activity measured by pairs 1 -2 and 2—4, and also between 1-3 and 3 -4 , i.e. was the EMG activity the same on the m. orb. oris superior and m. orb. oris inferior? (3) Whilst accepting inter-subject variation, was there any consistency in the results obtained from each subject throughout a series of experimental runs? (4) Using many repetitions of the utterance /pip/ could any significant differences be measured between the EMG activity resulting from the initial /p/ and the final / p/?
Method
Downloaded by: King's College London 137.73.144.138 - 3/6/2018 9:04:34 PM
The skin of the subject was suitably prepared, and four standard silver/silver chloride cup electrodes (having a 5 mm active diameter) were attached using adhesive discs, in the positions shown in figure 1A. Each electrode position was as near as possible to the vermil ion border to try to limit the activity measured to that of the m. orb. oris. The electrodes were filled with standard electrode jelly, and did not impede the normal movements of the lips. A fifth electrode was attached to the subject’s forehead to act as a ground electrode. The electrical activity measured between the electrode pairs was amplified using four biological amplifiers having identical frequency characteristics, namely, flat in the range 1.5-700 Hz. These signals were then recorded on four channels of an Ampex SP300 FM tape recorder, run at a tape speed of 3.75 ips to give a flat frequency response from DC to 625 Hz. The gains of each channel were set equal, and the channels monitored in turn using
EMG of Muscle Orbicularis Oris
179
Downloaded by: King's College London 137.73.144.138 - 3/6/2018 9:04:34 PM
a Tektronix 502A oscilloscope, to ensure good electrode contact, and a lack of spurious electrical pick-up. The subject’s audio output was obtained using a throat contact micro phone and recorded on another channel of the Ampex recorder, used in direct record mode. The subject was asked to repeat /pip/ at 5-scc intervals, in as nearly as possible his normal voice. The analysis system is shown diagramatically in figure 2. Because an averaging tech nique was used to eliminate individual variations, and give a result indicating the general trend, it was necessary to obtain a consistent trigger signal to control which part of the EMG signal was sampled. Also, because the trigger was derived from the audio signal, which started later than the EMG activity, it was necessary to delay the EMG signal with respect to the audio output, so that the trigger signal arrived at the averager slightly before the processed EMG signal (1). The EMG signal was re-recorded on another FM tape recorder and the delay introduced between the record and playback heads amounted to 0.45 sec. This FM recorder, which was based on a Ferrograph tape deck, ran at a tape speed of 3.75 ips, giving a flat frequency response from DC to 625 Hz. The EMG signal was rectified using a full-wave bridge rectifier and then integrated, the time constant of the integrator being 50 msec. This rectified and integrated EMG signal was fed into the input of the signal averager (Medelec AVM 62). This had a 200 point store and a resolution of 5 msec with a 1 sec averaging time. Tire trigger signal for the averager was obtained by First rectifying and integrating the audio signal using identical circuitry to that used for the EMG signal. This was then passed to a Schmitt trigger circuit, which had a controllable dead time, during which it could not be retriggered. This dead time was set at 4 sec, and thus prevented unrelated microphone pick-up from producing spurious triggers in the averager, which was triggered directly by the Schmitt output pulse. During playback, the Medelec was used in 'superposition’ mode, so that the outputs of the individual utterances were superimposed, to give an indication of the variation present. A permanent record of this and also the total average was obtained using the built-in U-V recorder of the instrument. A record of the final average was also obtained on a larger scale using a Bryans Auto Plotter, 22000 series. The averages of each channel could thus be superimposed to make qualitative and quantitative measurements easier. The individual EMG signals, even when rectified and integrated, were still very complex, there being considerable variation in the fine structure between utterances (fig. 3). However, the main peaks were consistently evident, and so the averaging technique was valid to eliminate the small random fluctuations, and leave a result representative of the general trend (fig. 4). In order to determine the size of a representative average, several trial runs were performed, during which 80 utterances were recorded. The averages of various samples of 30 were obtained, and there was found to be no significant difference between any of these. Thus the figure of 30 was, chosen as producing a representative result. During the experiment, the subjects produced 40 repetitions of the utterance /pip/, and the middle 30 were used for sampling so as to avoid any artefacts in the first 5 and the last 5 tokens. To obtain a measure of the variation inherent in the method, due to slight differences in the production of the utterances, several sets of results were displayed individ ually (fig. 3) using a Mingograf 34T ink jet recorder. The means and standard deviations of the magnitudes of the various peaks were calculated. The percentage variations (SD/M) ranged from 13.3 to 22.1 % with a mean of 18.3 %. The experimental subjects were two normal British males aged 24 and 25 years. Be cause there were several individual differences in the results, these have been presented separately for each subject.
Kelman ¡Gatehouse
180
2
an"
ft
•nteg audio
il
EM C
emg
—
-
t SfC
3
Downloaded by: King's College London 137.73.144.138 - 3/6/2018 9:04:34 PM
Fig. 2. Schematic of analysis system. Fig. 3. Four examples of rectified and integrated audio and EMG signals, together with the unprocessed EMG signal.
EMG of Muscle Orbicularis Oris
181
Fig. 4. Four examples of superpositions and means of a series of 30 processed EMG signals.
= Superior;
= inferior. Downloaded by: King's College London 137.73.144.138 - 3/6/2018 9:04:34 PM
Fig. 5. Typical results obtained from subject A.
Kelman/Gatehouse
182
Left electrode pairs
Right electrode pa irs
16 -»
P2
P3
Left e le ctrod e p a irs
P4
R ig h t electrode pairs
Downloaded by: King's College London 137.73.144.138 - 3/6/2018 9:04:34 PM
PI
EMG of Muscle Orbicularis Oris
Left electrode pa irs
183
R ight electrode p a irs
9
Le ft electrode pa irs
Right electrode p a irs
Fig. 6. Magnitude distribution of the results from subject A. o = Superior electrode;« = inferior electrode. Fig. 7. Magnitude distribution of the normalised results from subject A. o = Superior electrode; • = inferior electrode. Fig. 8. Typical results obtained from subject B.------ = Superior;------- = inferior. Fig. 9. Magnitude distribution of results from subject B. o = Superior electrode; • = inferior electrode. Fig. 10. Magnitude distribution of normalised results from subject B. o = Superior electrode; • = inferior electrode.
Downloaded by: King's College London 137.73.144.138 - 3/6/2018 9:04:34 PM
10
Kelman/Gatehouse
184
Results Subject A A typical example of the results obtained from subject A is shown in fig ure 5, and the same nomenclature is used throughout. From the m. orb. oris superior, two main peaks PI and P2, relating to the initial /pi and final / p/ were observed. A much smaller peak, P3, was evident, lying between PI and P2. How ever, from the m. orb. oris, inferior, PI and P2 were much smaller and no longer dominant, P3 being of a similar magnitude. Another peak, P4, occurring after P2, was also observed. Figure 6 illustrates the results obtained for the magnitudes of the peaks from a series of experimental sessions. The means and standard deviations of these results were calculated for each peak, and left-right comparisons were carried out using Student’s t test and the Wilcoxon rank test. The results are tabulated in table 1. There was found to be no significant difference at the 0.05 level in the range of magnitude of any peak due to left-right asymmetry. However, when the results shown in figure 6 were paired according to re cording session, and a paired t test performed, the peaks PI and P2 from the right electrode pairs were found to be significantly larger, at the 0.05 level, than those from the corresponding left electrode pairs (table 1). There were no signif icant differences in the magnitudes of P3 or P4. A comparison was also made between the magnitudes of the same peaks obtained from superior and inferior electrode pairs (table II), and in each case there was a significant difference. To investigate whether or not there was a constant ratio of P2:P1 or P3:P1 from the superior and inferior electrode pairs, the magnitude of each PI was set as 10 units, and the normalised values of the other peaks calculated. The results are shown in figure 7, and again the left-right symmetry was maintained (table III). The result for P3 for the inferior pairs did not seem to be representa tive and was considered an artefact. For the superior—inferior comparison, although the ranges for P2 normal ised seemed very close, both tests indicated a significant difference at the 0.05 level. The test values for P3 showed that there was an even more significant difference between the superior and inferior muscles in the EMG activity for this peak than for P2. These results will be discussed more fully later.
Downloaded by: King's College London 137.73.144.138 - 3/6/2018 9:04:34 PM
Subject B The general form of the results obtained from subject B is shown in figure 8. From the m. orb. oris, superior, PI and P2 were dominant with P3 noticeable, but rarely large enough to be measurable. From the m. orb. oris, inferior, PI and P2 were again much smaller with P3 increased in magnitude. The later peak was also found, and as this was judged to correspond to P4 in subject A, was also named P4. However, the timing of the late peak from the superior muscle was
liMG of Muscle Orbicularis Oris
185
Table I. Statistical tests on results from subject A for left-right comparison Peak
Electrode pairs
n
t value
Significance at 0.05 level
Wilcoxon coeff.
Significance at 0.05 level
PI PI P2 P2 P3 P3 P4
1 -3 /1 -2 3 4/2-4 1-3/1 -2 3 4/2-4 1—3/1 —2 3—4/2-4 3 -4 /2 -4
38 37 38 37 38 37 32
0.50 1.30 0.38 0.51 0.29 0.76 0.05
NS NS NS NS NS NS NS
453.5 469 444 430 447.5 365.5 286.5
NS NS NS NS NS NS NS
18 18 18 18 18 18 14
4.23 4.36 2.95 3.75 0.93 1.15 0.39
S
Paired lest PI PI P2 P2 P3 P3 P4
1- 3 /1 -2 3 —4/2—4 1—3/1—2 3 - 4 /2 -4 1- 3 /1 -2 3 -4 /2 -4 3—4/2—4
s s s NS NS NS
Peak
Electrode pairs
n
t value
Significance at 0.05 level
Wilcoxon coeff.
Significance at 0.05 level
PI PI P2 P2 P3 P3
1 -3 /3 -4 1 -2 /2 -4 1 -3 /3 -4 1-2 /2 -4 1 -3 /3 -4 1 -2 /2 -4
39 36 39 36 39 36
7.75 8.28 10.55 8.88 3.79 6.68
S S S S S S
635.5 558.5 610 610 321 197
S S S S S S
significantly different from that of P4, and so the late peak from the superior electrode pairs was designated P5. The data obtained from a series of experimental sessions are displayed in figure 9. Again the t test and rank test were carried out on the ranges obtained for corresponding peaks from left and right electrode pairs (table IV). No signif icant differences, at the 0.05 level, were obtained for the ranges of any peak. When the paired test was carried out, there was found to be no consistent difference in peak size from left to right except for peak P2 from the inferior electrode pairs. It was concluded that in general, there was no consistent le ftright asymmetry for this subject, apart perhaps for P2 for m. orb. oris, inferior.
Downloaded by: King's College London 137.73.144.138 - 3/6/2018 9:04:34 PM
Table II. Statistical tests on results from subject A for superior-inferior comparison
186
Kelman/Gatehouse Table III. Statistical tests on normalised results from subject A Peak
Electrode pairs
n
t value
PI P2 P3 P3 P4
1 3 /1 -2 3 -4 /2 —4 1-3/1 -2 3 - 4 /2 -4 3-4/2 4
39 37 39 36 32
0.28 0.62 0.03 2.50 1.06
P2 P2 P3 P3
1-3/3 4 1 2/2-4 1 -3 /3 -4 1 -2 /2 -4
40 36 40 35
4.00 2.82 10.00 23.49
Significance at 0.05 level
Wilcoxon coeff.
Significance at 0.05 level
NS NS NS S NS
438.5 375 446 298.5 223
NS NS NS S NS
S S S S
612 567.5 253 190
S
s s s
Table IV. Statistical tests on results from subject B for left -right comparison Peak
Electrode pairs
n
t value
Significance at 0.05 level
Wilcoxon coeff.
Significance at 0.05 level
PI PI P2 P2 P3 P4 PS
1-3/1 2 3 - 4 /2 -4 1—3/1—2 3—4/2 4 3 -4 /2 -4 3 - 4 /2 -4 1—3/1 —2
32 25 32 30 30 29 26
0.18 0.62 1.42 1.25 0.45 1.10 0.80
NS NS NS NS NS NS NS
339 152.5 265 200 261 265 256
NS NS NS NS NS NS NS
15 12 15 14 14 14 9
0.73 1.31 3.43 2.50 0.67 1.11 1.58
NS NS S S NS NS NS
Paired test PI PI P2 P2 P3 P4 P5
1—3/1 —2 3 —4/2—4 1—3/1 —2 3 -4 /2 -4 3 -4 /2 -4 3—4/2—4 1 —3/1—2
Downloaded by: King's College London 137.73.144.138 - 3/6/2018 9:04:34 PM
The t tests were also carried out to compare the magnitudes of corre sponding peaks obtained from the superior and inferior electrode pairs. Signif icant differences were obtained for peaks PI and P2 (table V), but peaks P3, P4 and P5 could not be compared as they were only evident on either the superior or inferior muscle. The absolute magnitudes of the peak heights were then normalised, again setting the value of each PI at 10 units, and the results plotted (fig. 10). No consistent left—right asymmetries were found, and the ranges of P2
EMG of Muscle Orbicularis Oris
187
Table V. Statistical tests on results from subject B for superior-inferior comparison Peak
Electrode pairs
n
t value
Significance at 0.05 level
Wilcoxon coeff.
Significance at 0.05 level
PI PI P2 P2
1—3/3 - 4 1 -2 /2 -4 1 -3 /3 -4 1- 2/2-4
29 28 32 30
3.44 2.43 4.05 5.04
S S
351.5 310 396 358.5
S S S S
s s
Table VI. Statistical tests on normalised results for subject B Peak
Electrode pairs
n
t value
Significance at 0.05 level
Wilcoxon coeff.
Significance at 0.05 level
P2 P2 P3 P4 P5
1- 3 /1 -2 3 -4/2—4 3—4/2—4 3 -4 /2 -4 1-3/1 2
32 25 25 25 25
1.65 2.54 0.54 1.65 0.56
NS S NS NS NS
284.5 134 206 214 242
NS S NS NS NS
P2 P2
1 -3 /3 -4 1 -2 /2 -4
29 28
1.90 2.39
NS S
319.5 311
NS S
Electrode pair
n
t value
Significance at 0.05 level
Wilcoxon coeff.
Significance at 0.05 level
Subject A 1-3 1-2 3 -4 2-4
40 36 38 36
0.80 0.56 3.46 2.66
NS NS S S
476.5 387.5 547.5 459
NS NS S S
Subject B 1-3 1-2 3-4 2-4
34 30 27 28
0.70 2.00 0.58 0.00
NS NS NS NS
336.5 198.5 221 224.5
NS S NS NS Downloaded by: King's College London 137.73.144.138 - 3/6/2018 9:04:34 PM
Table VII. Statistical tests on initial and final /p/’s
Kelman /Gatehouse
188
normalised for all four electrode pairs were very close. Comparison between the superior and inferior EMG activity was only possible for peak P2, and the results were inconclusive (table VI). Initial and Final /p/ Using the data previously described, the magnitudes of PI and P2 were compared (table VII). For subject B, there was found to be no generally signif icant difference for any of the electrode pairs. For subject A, there was no significant difference for the superior electrode pairs, but both inferior pairs showed significant differences at the 0.05 level.
There was surprisingly little variation from run to run on each subject, and so a reasonable consistency could be achieved. In general, there were found to be no significant left-right differences in the ranges of peak magnitudes resulting from a series of experimental sessions. However, for subject A, when a paired t test was carried out, there seemed to be significantly greater EMG activity from the right electrode pairs than from the left. This appeared to be a characteristic of the subject. The results obtained from m. orb. oris, superior and m. orb. oris, inferior were markedly different. This seems to agree qualitatively with the results of Leanderson et al. (3), who, although using needle electrodes and VCV utter ances, also appeared to get different activity from the superior and inferior muscles. In agreement with Fromkin (1) and Tatliam and Morton (10), who used different surface electrode configurations, there was little difference between the EMG activity produced for the initial and final /p/ in the utterance /pip/. Although the general forms of the EMG activity were fairly similar for both subjects, there were also obvious differences, which demonstrated inter-subject variability. It is clear that a similar investigation on many more subjects could be of value. The possibility of contributions from muscles other than the m. orb. oris could not be excluded as surface electrodes were used in this experiment; there fore, no definite conclusions could be drawn as to the origin of the various peaks. However, the use of different electrode configurations, such as those shown in figure IB and C, may provide a further insight into this problem. The experiment shows the important fact that the EMG activities from the m. orb. oris, superior and m. orb. oris, inferior are quite different. It seems evident that the exact electrode position on either muscle, and also electrode configuration will be equally important, and it is hoped to report on these factors in future articles.
Downloaded by: King's College London 137.73.144.138 - 3/6/2018 9:04:34 PM
Conclusions
EMG of Muscle Orbicularis Oris
189
Summary Surface electrodes were used in a study of the EMG activity of the m. orb. oris on two subjects during phonation of the CVC utterance /pip/. An averaging technique was em ployed to facilitate comparison between the activity from different electrode pairs. The results were statistically analysed to measure the significance of any differences obtained.
A cknowledgement The authors would like to thank Prof. J.M.A. Lenihan of the Department of Clinical Physics and Bio-Engineering for his support during this investigation.
References 1 Fromkin, V.: Neuro-muscular specification of linguistic units. Lang. Speech 9: 170 199 (1966). 2 Fromkin, V. and Ladefoged, P.: Electromyography in speech research. Phonetica 15: 219-242 (1966). 3 Leanderson, R. and Lindhlom, B.F.F.: Muscle activation for labial speech gestures. Acta oto-lar. 73: 36 2 - 373 (19 72). 4 Leanderson, R.: Persson, A., and Oilman, S.: Electromyographic studies of the func tion of the facial muscles in dysarthria. Acta oto-lar. 71: 89 94 (1970). 5 Leanderson, R.: Persson, A., and Oilman, S.: Electromyographic studies of facial mus cle activity in speech. Acta oto-lar. 72: 361-369 (1971). 6 Lubker, J.F. and Parris, P.J.: Simultaneous measurements of intraoral pressure, force of labial contact, and labial electromyographic activity during production of the stop consonant cognates/p/ and /b/. J. acoust. soc. Am. 47: 625-633 (1970). 7 MacNeilage, P.F.: Electromyographic and acoustic study of the production of certain final clusters. J. acoust. soc. Am. 35: 461-463 (1963). 8 MacNeilage, P.F. and DeClerk, J.L.: On the motor control of coarticulation in CVC monosyllables. J. acoust. soc. Am. 45: 1217, 1233 (1969). 9 Tatham, M.A.A. and Morton, K.: Some electromyography data towards a model of speech production. Lang. Speech 12: 39-53 (1969). 10 Tatham. M.A.A. and Morton, K.: Electromyographic and intraoral air-pressure studies of bi-labial stops. Lang. Speech 16: 336-350 (1970).
Downloaded by: King's College London 137.73.144.138 - 3/6/2018 9:04:34 PM
Mr. A.W. Kelman, Department of Clinical Physics and Bio-Engineering, West of Scotland Health Boards, 11 West Graham Street, Glasgow, G4: 9LF (UK)
A Study of the Electromyographic Activity of the Muscle Orbicularis Oris A. W. Kelman and S. Gatehouse West of Scotland Health Boards, Department of Clinical Physics and Bio-Engineering, Glasgow
Introduction The relevance of electromyographic (EMG) data to phonetics has already been recognised, and the EMG activity of the facial muscles, while the subject phonated a series of consonant-vowel-consonant (CVC) or vowel-consonantvowel (VCV) utterances has been investigated (1, 2, 6 -9 ). In one series of experiments, Leanderson el al. (4, 5) investigated several facial muscles simulta neously, using needle electrodes. However, most of the research has utilised surface electrodes placed on the muscle orbicularis oris (m. orb. oris). Since different electrode sites and configurations have been used, e.g. upper lip or lower lip, on midline or off midline, it is very difficult to correlate results obtained from different experiments. Accepting the fact that surface electrodes measure activity from the surrounding muscle and not from a single or small group of motor units, it seemed evident that a study of the EMG output asso ciated with different electrode positions was required. A series of experiments was thus undertaken in which the EMG activity could be measured simultaneously from four electrode sites in the configurations A, B and C (fig. 1). The electrodes could be connected in either differential (bipolar) pairs or in a monopolar mode, with the reference electrode situated on the mastoid. The subject was asked to repeat many times the utterances: Ip Ip/, /pAp/, /plb/, /bib/, /blp/, /bAb/. These were chosen so as to investigate whether there were any significant differences in the EMG activity produced by the m. orb. oris for the labial sounds / p/ and /b/, in different phonetic contexts.
Downloaded by: King's College London 137.73.144.138 - 3/6/2018 9:04:34 PM
Received: February 7, 1975; accepted: March 9, 1975.
Kelman/Gatehouse
178
In this paper we report on the first experiment in which the utterance /pip/ was used and where the objectives were much more limited, namely, using the electrode configurations shown in figure 1 A with the electrodes connected in differential mode in the sequence 1-2, 1-3 , 2 -4 , and 3 -4 . (1 ) Were there any significant differences between the EMG activity mea sured by electrode pairs 1 -2 and 1 -3 , and also between 2 - 4 and 3 -4 , i.e. was there left-right symmetry about the vertical midline? (2) Were there any significant differences between the activity measured by pairs 1 -2 and 2—4, and also between 1-3 and 3 -4 , i.e. was the EMG activity the same on the m. orb. oris superior and m. orb. oris inferior? (3) Whilst accepting inter-subject variation, was there any consistency in the results obtained from each subject throughout a series of experimental runs? (4) Using many repetitions of the utterance /pip/ could any significant differences be measured between the EMG activity resulting from the initial /p/ and the final / p/?
Method
Downloaded by: King's College London 137.73.144.138 - 3/6/2018 9:04:34 PM
The skin of the subject was suitably prepared, and four standard silver/silver chloride cup electrodes (having a 5 mm active diameter) were attached using adhesive discs, in the positions shown in figure 1A. Each electrode position was as near as possible to the vermil ion border to try to limit the activity measured to that of the m. orb. oris. The electrodes were filled with standard electrode jelly, and did not impede the normal movements of the lips. A fifth electrode was attached to the subject’s forehead to act as a ground electrode. The electrical activity measured between the electrode pairs was amplified using four biological amplifiers having identical frequency characteristics, namely, flat in the range 1.5-700 Hz. These signals were then recorded on four channels of an Ampex SP300 FM tape recorder, run at a tape speed of 3.75 ips to give a flat frequency response from DC to 625 Hz. The gains of each channel were set equal, and the channels monitored in turn using
EMG of Muscle Orbicularis Oris
179
Downloaded by: King's College London 137.73.144.138 - 3/6/2018 9:04:34 PM
a Tektronix 502A oscilloscope, to ensure good electrode contact, and a lack of spurious electrical pick-up. The subject’s audio output was obtained using a throat contact micro phone and recorded on another channel of the Ampex recorder, used in direct record mode. The subject was asked to repeat /pip/ at 5-scc intervals, in as nearly as possible his normal voice. The analysis system is shown diagramatically in figure 2. Because an averaging tech nique was used to eliminate individual variations, and give a result indicating the general trend, it was necessary to obtain a consistent trigger signal to control which part of the EMG signal was sampled. Also, because the trigger was derived from the audio signal, which started later than the EMG activity, it was necessary to delay the EMG signal with respect to the audio output, so that the trigger signal arrived at the averager slightly before the processed EMG signal (1). The EMG signal was re-recorded on another FM tape recorder and the delay introduced between the record and playback heads amounted to 0.45 sec. This FM recorder, which was based on a Ferrograph tape deck, ran at a tape speed of 3.75 ips, giving a flat frequency response from DC to 625 Hz. The EMG signal was rectified using a full-wave bridge rectifier and then integrated, the time constant of the integrator being 50 msec. This rectified and integrated EMG signal was fed into the input of the signal averager (Medelec AVM 62). This had a 200 point store and a resolution of 5 msec with a 1 sec averaging time. Tire trigger signal for the averager was obtained by First rectifying and integrating the audio signal using identical circuitry to that used for the EMG signal. This was then passed to a Schmitt trigger circuit, which had a controllable dead time, during which it could not be retriggered. This dead time was set at 4 sec, and thus prevented unrelated microphone pick-up from producing spurious triggers in the averager, which was triggered directly by the Schmitt output pulse. During playback, the Medelec was used in 'superposition’ mode, so that the outputs of the individual utterances were superimposed, to give an indication of the variation present. A permanent record of this and also the total average was obtained using the built-in U-V recorder of the instrument. A record of the final average was also obtained on a larger scale using a Bryans Auto Plotter, 22000 series. The averages of each channel could thus be superimposed to make qualitative and quantitative measurements easier. The individual EMG signals, even when rectified and integrated, were still very complex, there being considerable variation in the fine structure between utterances (fig. 3). However, the main peaks were consistently evident, and so the averaging technique was valid to eliminate the small random fluctuations, and leave a result representative of the general trend (fig. 4). In order to determine the size of a representative average, several trial runs were performed, during which 80 utterances were recorded. The averages of various samples of 30 were obtained, and there was found to be no significant difference between any of these. Thus the figure of 30 was, chosen as producing a representative result. During the experiment, the subjects produced 40 repetitions of the utterance /pip/, and the middle 30 were used for sampling so as to avoid any artefacts in the first 5 and the last 5 tokens. To obtain a measure of the variation inherent in the method, due to slight differences in the production of the utterances, several sets of results were displayed individ ually (fig. 3) using a Mingograf 34T ink jet recorder. The means and standard deviations of the magnitudes of the various peaks were calculated. The percentage variations (SD/M) ranged from 13.3 to 22.1 % with a mean of 18.3 %. The experimental subjects were two normal British males aged 24 and 25 years. Be cause there were several individual differences in the results, these have been presented separately for each subject.
Kelman ¡Gatehouse
180
2
an"
ft
•nteg audio
il
EM C
emg
—
-
t SfC
3
Downloaded by: King's College London 137.73.144.138 - 3/6/2018 9:04:34 PM
Fig. 2. Schematic of analysis system. Fig. 3. Four examples of rectified and integrated audio and EMG signals, together with the unprocessed EMG signal.
EMG of Muscle Orbicularis Oris
181
Fig. 4. Four examples of superpositions and means of a series of 30 processed EMG signals.
= Superior;
= inferior. Downloaded by: King's College London 137.73.144.138 - 3/6/2018 9:04:34 PM
Fig. 5. Typical results obtained from subject A.
Kelman/Gatehouse
182
Left electrode pairs
Right electrode pa irs
16 -»
P2
P3
Left e le ctrod e p a irs
P4
R ig h t electrode pairs
Downloaded by: King's College London 137.73.144.138 - 3/6/2018 9:04:34 PM
PI
EMG of Muscle Orbicularis Oris
Left electrode pa irs
183
R ight electrode p a irs
9
Le ft electrode pa irs
Right electrode p a irs
Fig. 6. Magnitude distribution of the results from subject A. o = Superior electrode;« = inferior electrode. Fig. 7. Magnitude distribution of the normalised results from subject A. o = Superior electrode; • = inferior electrode. Fig. 8. Typical results obtained from subject B.------ = Superior;------- = inferior. Fig. 9. Magnitude distribution of results from subject B. o = Superior electrode; • = inferior electrode. Fig. 10. Magnitude distribution of normalised results from subject B. o = Superior electrode; • = inferior electrode.
Downloaded by: King's College London 137.73.144.138 - 3/6/2018 9:04:34 PM
10
Kelman/Gatehouse
184
Results Subject A A typical example of the results obtained from subject A is shown in fig ure 5, and the same nomenclature is used throughout. From the m. orb. oris superior, two main peaks PI and P2, relating to the initial /pi and final / p/ were observed. A much smaller peak, P3, was evident, lying between PI and P2. How ever, from the m. orb. oris, inferior, PI and P2 were much smaller and no longer dominant, P3 being of a similar magnitude. Another peak, P4, occurring after P2, was also observed. Figure 6 illustrates the results obtained for the magnitudes of the peaks from a series of experimental sessions. The means and standard deviations of these results were calculated for each peak, and left-right comparisons were carried out using Student’s t test and the Wilcoxon rank test. The results are tabulated in table 1. There was found to be no significant difference at the 0.05 level in the range of magnitude of any peak due to left-right asymmetry. However, when the results shown in figure 6 were paired according to re cording session, and a paired t test performed, the peaks PI and P2 from the right electrode pairs were found to be significantly larger, at the 0.05 level, than those from the corresponding left electrode pairs (table 1). There were no signif icant differences in the magnitudes of P3 or P4. A comparison was also made between the magnitudes of the same peaks obtained from superior and inferior electrode pairs (table II), and in each case there was a significant difference. To investigate whether or not there was a constant ratio of P2:P1 or P3:P1 from the superior and inferior electrode pairs, the magnitude of each PI was set as 10 units, and the normalised values of the other peaks calculated. The results are shown in figure 7, and again the left-right symmetry was maintained (table III). The result for P3 for the inferior pairs did not seem to be representa tive and was considered an artefact. For the superior—inferior comparison, although the ranges for P2 normal ised seemed very close, both tests indicated a significant difference at the 0.05 level. The test values for P3 showed that there was an even more significant difference between the superior and inferior muscles in the EMG activity for this peak than for P2. These results will be discussed more fully later.
Downloaded by: King's College London 137.73.144.138 - 3/6/2018 9:04:34 PM
Subject B The general form of the results obtained from subject B is shown in figure 8. From the m. orb. oris, superior, PI and P2 were dominant with P3 noticeable, but rarely large enough to be measurable. From the m. orb. oris, inferior, PI and P2 were again much smaller with P3 increased in magnitude. The later peak was also found, and as this was judged to correspond to P4 in subject A, was also named P4. However, the timing of the late peak from the superior muscle was
liMG of Muscle Orbicularis Oris
185
Table I. Statistical tests on results from subject A for left-right comparison Peak
Electrode pairs
n
t value
Significance at 0.05 level
Wilcoxon coeff.
Significance at 0.05 level
PI PI P2 P2 P3 P3 P4
1 -3 /1 -2 3 4/2-4 1-3/1 -2 3 4/2-4 1—3/1 —2 3—4/2-4 3 -4 /2 -4
38 37 38 37 38 37 32
0.50 1.30 0.38 0.51 0.29 0.76 0.05
NS NS NS NS NS NS NS
453.5 469 444 430 447.5 365.5 286.5
NS NS NS NS NS NS NS
18 18 18 18 18 18 14
4.23 4.36 2.95 3.75 0.93 1.15 0.39
S
Paired lest PI PI P2 P2 P3 P3 P4
1- 3 /1 -2 3 —4/2—4 1—3/1—2 3 - 4 /2 -4 1- 3 /1 -2 3 -4 /2 -4 3—4/2—4
s s s NS NS NS
Peak
Electrode pairs
n
t value
Significance at 0.05 level
Wilcoxon coeff.
Significance at 0.05 level
PI PI P2 P2 P3 P3
1 -3 /3 -4 1 -2 /2 -4 1 -3 /3 -4 1-2 /2 -4 1 -3 /3 -4 1 -2 /2 -4
39 36 39 36 39 36
7.75 8.28 10.55 8.88 3.79 6.68
S S S S S S
635.5 558.5 610 610 321 197
S S S S S S
significantly different from that of P4, and so the late peak from the superior electrode pairs was designated P5. The data obtained from a series of experimental sessions are displayed in figure 9. Again the t test and rank test were carried out on the ranges obtained for corresponding peaks from left and right electrode pairs (table IV). No signif icant differences, at the 0.05 level, were obtained for the ranges of any peak. When the paired test was carried out, there was found to be no consistent difference in peak size from left to right except for peak P2 from the inferior electrode pairs. It was concluded that in general, there was no consistent le ftright asymmetry for this subject, apart perhaps for P2 for m. orb. oris, inferior.
Downloaded by: King's College London 137.73.144.138 - 3/6/2018 9:04:34 PM
Table II. Statistical tests on results from subject A for superior-inferior comparison
186
Kelman/Gatehouse Table III. Statistical tests on normalised results from subject A Peak
Electrode pairs
n
t value
PI P2 P3 P3 P4
1 3 /1 -2 3 -4 /2 —4 1-3/1 -2 3 - 4 /2 -4 3-4/2 4
39 37 39 36 32
0.28 0.62 0.03 2.50 1.06
P2 P2 P3 P3
1-3/3 4 1 2/2-4 1 -3 /3 -4 1 -2 /2 -4
40 36 40 35
4.00 2.82 10.00 23.49
Significance at 0.05 level
Wilcoxon coeff.
Significance at 0.05 level
NS NS NS S NS
438.5 375 446 298.5 223
NS NS NS S NS
S S S S
612 567.5 253 190
S
s s s
Table IV. Statistical tests on results from subject B for left -right comparison Peak
Electrode pairs
n
t value
Significance at 0.05 level
Wilcoxon coeff.
Significance at 0.05 level
PI PI P2 P2 P3 P4 PS
1-3/1 2 3 - 4 /2 -4 1—3/1—2 3—4/2 4 3 -4 /2 -4 3 - 4 /2 -4 1—3/1 —2
32 25 32 30 30 29 26
0.18 0.62 1.42 1.25 0.45 1.10 0.80
NS NS NS NS NS NS NS
339 152.5 265 200 261 265 256
NS NS NS NS NS NS NS
15 12 15 14 14 14 9
0.73 1.31 3.43 2.50 0.67 1.11 1.58
NS NS S S NS NS NS
Paired test PI PI P2 P2 P3 P4 P5
1—3/1 —2 3 —4/2—4 1—3/1 —2 3 -4 /2 -4 3 -4 /2 -4 3—4/2—4 1 —3/1—2
Downloaded by: King's College London 137.73.144.138 - 3/6/2018 9:04:34 PM
The t tests were also carried out to compare the magnitudes of corre sponding peaks obtained from the superior and inferior electrode pairs. Signif icant differences were obtained for peaks PI and P2 (table V), but peaks P3, P4 and P5 could not be compared as they were only evident on either the superior or inferior muscle. The absolute magnitudes of the peak heights were then normalised, again setting the value of each PI at 10 units, and the results plotted (fig. 10). No consistent left—right asymmetries were found, and the ranges of P2
EMG of Muscle Orbicularis Oris
187
Table V. Statistical tests on results from subject B for superior-inferior comparison Peak
Electrode pairs
n
t value
Significance at 0.05 level
Wilcoxon coeff.
Significance at 0.05 level
PI PI P2 P2
1—3/3 - 4 1 -2 /2 -4 1 -3 /3 -4 1- 2/2-4
29 28 32 30
3.44 2.43 4.05 5.04
S S
351.5 310 396 358.5
S S S S
s s
Table VI. Statistical tests on normalised results for subject B Peak
Electrode pairs
n
t value
Significance at 0.05 level
Wilcoxon coeff.
Significance at 0.05 level
P2 P2 P3 P4 P5
1- 3 /1 -2 3 -4/2—4 3—4/2—4 3 -4 /2 -4 1-3/1 2
32 25 25 25 25
1.65 2.54 0.54 1.65 0.56
NS S NS NS NS
284.5 134 206 214 242
NS S NS NS NS
P2 P2
1 -3 /3 -4 1 -2 /2 -4
29 28
1.90 2.39
NS S
319.5 311
NS S
Electrode pair
n
t value
Significance at 0.05 level
Wilcoxon coeff.
Significance at 0.05 level
Subject A 1-3 1-2 3 -4 2-4
40 36 38 36
0.80 0.56 3.46 2.66
NS NS S S
476.5 387.5 547.5 459
NS NS S S
Subject B 1-3 1-2 3-4 2-4
34 30 27 28
0.70 2.00 0.58 0.00
NS NS NS NS
336.5 198.5 221 224.5
NS S NS NS Downloaded by: King's College London 137.73.144.138 - 3/6/2018 9:04:34 PM
Table VII. Statistical tests on initial and final /p/’s
Kelman /Gatehouse
188
normalised for all four electrode pairs were very close. Comparison between the superior and inferior EMG activity was only possible for peak P2, and the results were inconclusive (table VI). Initial and Final /p/ Using the data previously described, the magnitudes of PI and P2 were compared (table VII). For subject B, there was found to be no generally signif icant difference for any of the electrode pairs. For subject A, there was no significant difference for the superior electrode pairs, but both inferior pairs showed significant differences at the 0.05 level.
There was surprisingly little variation from run to run on each subject, and so a reasonable consistency could be achieved. In general, there were found to be no significant left-right differences in the ranges of peak magnitudes resulting from a series of experimental sessions. However, for subject A, when a paired t test was carried out, there seemed to be significantly greater EMG activity from the right electrode pairs than from the left. This appeared to be a characteristic of the subject. The results obtained from m. orb. oris, superior and m. orb. oris, inferior were markedly different. This seems to agree qualitatively with the results of Leanderson et al. (3), who, although using needle electrodes and VCV utter ances, also appeared to get different activity from the superior and inferior muscles. In agreement with Fromkin (1) and Tatliam and Morton (10), who used different surface electrode configurations, there was little difference between the EMG activity produced for the initial and final /p/ in the utterance /pip/. Although the general forms of the EMG activity were fairly similar for both subjects, there were also obvious differences, which demonstrated inter-subject variability. It is clear that a similar investigation on many more subjects could be of value. The possibility of contributions from muscles other than the m. orb. oris could not be excluded as surface electrodes were used in this experiment; there fore, no definite conclusions could be drawn as to the origin of the various peaks. However, the use of different electrode configurations, such as those shown in figure IB and C, may provide a further insight into this problem. The experiment shows the important fact that the EMG activities from the m. orb. oris, superior and m. orb. oris, inferior are quite different. It seems evident that the exact electrode position on either muscle, and also electrode configuration will be equally important, and it is hoped to report on these factors in future articles.
Downloaded by: King's College London 137.73.144.138 - 3/6/2018 9:04:34 PM
Conclusions
EMG of Muscle Orbicularis Oris
189
Summary Surface electrodes were used in a study of the EMG activity of the m. orb. oris on two subjects during phonation of the CVC utterance /pip/. An averaging technique was em ployed to facilitate comparison between the activity from different electrode pairs. The results were statistically analysed to measure the significance of any differences obtained.
A cknowledgement The authors would like to thank Prof. J.M.A. Lenihan of the Department of Clinical Physics and Bio-Engineering for his support during this investigation.
References 1 Fromkin, V.: Neuro-muscular specification of linguistic units. Lang. Speech 9: 170 199 (1966). 2 Fromkin, V. and Ladefoged, P.: Electromyography in speech research. Phonetica 15: 219-242 (1966). 3 Leanderson, R. and Lindhlom, B.F.F.: Muscle activation for labial speech gestures. Acta oto-lar. 73: 36 2 - 373 (19 72). 4 Leanderson, R.: Persson, A., and Oilman, S.: Electromyographic studies of the func tion of the facial muscles in dysarthria. Acta oto-lar. 71: 89 94 (1970). 5 Leanderson, R.: Persson, A., and Oilman, S.: Electromyographic studies of facial mus cle activity in speech. Acta oto-lar. 72: 361-369 (1971). 6 Lubker, J.F. and Parris, P.J.: Simultaneous measurements of intraoral pressure, force of labial contact, and labial electromyographic activity during production of the stop consonant cognates/p/ and /b/. J. acoust. soc. Am. 47: 625-633 (1970). 7 MacNeilage, P.F.: Electromyographic and acoustic study of the production of certain final clusters. J. acoust. soc. Am. 35: 461-463 (1963). 8 MacNeilage, P.F. and DeClerk, J.L.: On the motor control of coarticulation in CVC monosyllables. J. acoust. soc. Am. 45: 1217, 1233 (1969). 9 Tatham, M.A.A. and Morton, K.: Some electromyography data towards a model of speech production. Lang. Speech 12: 39-53 (1969). 10 Tatham. M.A.A. and Morton, K.: Electromyographic and intraoral air-pressure studies of bi-labial stops. Lang. Speech 16: 336-350 (1970).
Downloaded by: King's College London 137.73.144.138 - 3/6/2018 9:04:34 PM
Mr. A.W. Kelman, Department of Clinical Physics and Bio-Engineering, West of Scotland Health Boards, 11 West Graham Street, Glasgow, G4: 9LF (UK)
