Dept. for Speech, Music and Hearing
Quarterly Progress and
Status Report
Spectrum factors relevant to
phonetogram measurement
Gramming, P. and Sundberg, J.
journal:
volume:
number:
year:
pages:
STL-QPSR
28
2-3
1987
039-061
http://www.speech.kth.se/qpsr
STL-QPSR 2-3/1987
111. MUSIC AculsTICS
SPECTRUM FACXORS FEIJWAN'I' TO PHONETOGRAM
Patricia Gramning* and Johan Sundberg
A.
Abst ract
Spectra of the vowel /a/ from tape-recorded phonetogram measurements of 20 healthy, untrained voices were analyzed. The overall SPL
was calculated from the loudest spectrum partials using both a f l a t a d
an A-weighted frequency curve. If a f l a t frguency curve was used, the
fundamental was generally the loudest partial in soft phonation throughout the fundamental frequency range, while in loud phonation an overtone
had the greatest amplitude. This suggests the possibility of a physiological interpret at ion of phonetograms i n terms of voice source characterist ics. The SPL d i f f e r e n c e between d i f f e r e n t vowels produced by 2 2
female speech therapist students seemed mainly dependent on the differences in the formant frequencies between the vowels. For a given vowel,
the f i r s t formant frequency was found generally t o be lower i n soft than
in loud phonation. It is concluded that an interpretat ion of a phonetogram in terms of voice physiology is somewhat easier i f a vowel with a
high f i r s t formant frequency is used, o r i f an A-weighted rather than a
f l a t frequency curve i s applied for the recordings.
Int roduct ion
A phonetogram shows the sound pressure level (SPL) of softest and
loudest possible phonat ion over the ent i r e fundamental frequency range
of a voice. I n a clinical voice analysis, an important purpose is t o
obtain information on the physiology of the vocal fold vibrations. The
SPL produced during phonat ion reflects certain aspects of these vibrat ions. Thus, a phonetogram has t h e potent i a l of r e f l e c t ing relevant
aspects of the larynqeal function of a voice (Klingholz & Elart in, 1983;
Schutte & ,Seidner, 1983; Stone & Ferch, 1982).
The peak amplitude of the flow qlottoaram (the t ransglottal airflow
waveform) i s c l o s e l y r e l a t e d t o t h e sound l e v e l of t h e lowest voice
source partial, or the fundamental; the negative peak amplitude of the
differentiated flow glottograrn i s decisive t o the amplitudes of the
voice source overtones (Gauffin & Sunclberg, 1980; Fant, Liljencrants, &
Lin, 1985).
*Deprtment of Phoniatrics, FX7T Clinic, Malmo qeneral Hospital, S-21401
Malmij, Sweden
STL-QPSR 2-3/1987
t i a l ) , the overall sound levels were calculated, applying a f l a t weighted curve. The correspxding levels i n dB(A) were obtained by applying
the A-weighted curve t o the levels of these same strongest spectrum
partials and then repeating the calculation using these new level values. These calculat ions were also checked by direct measurements. ?he
output of the tape recorder was connected t o a sound l e v e l recorder
(Briiel & Kjaer 2307), which was f i r s t adjusted t o a f l a t frequency curve
and then t o A-weighted frequency curve.
As the sound level often varied more or less during a vocalization,
the r e l i a b i l i t y of the spectrum analysis was checked. A l l vocalizations
prduced by two female subjects pertaining t o the f irst-ment ioned group
of subjects were independently analyzed three times. In the analysis of
each vocalization, t h e onset t r a n s i e n t was discarded. In t h i s way,
three level values were obtained for each spectrum partial appearing
below 5 kHz. The difference in dB between t h e extremes among these
three level values was used as an indicator of reproducibility.
Fig. l a shows the reproducibility of the level of the fundamental
for the two female subjects. The highest value of the difference for
both subject s was 1.7 dR in both soft and loud phonat ion, and in most
cases it was less than 1 dB. This reproducibility did not seem t o be
influenced neither by loudness of phonat ion, nor by fundamenta1 f requency. Fig. l b shows the corresponding values for the loudest spectrum
partial in loud phonat ion ( i n soft phonation the fundamental is almost
invariably the strongest part ial, as w i l l be demonstrated below).
A
maximum of 2.2 dB occurred at 440 Hz, but most values are smaller than 1
dB. I n Fig. lc, the values for the partial closest t o 1.5 kHz are shown.
In some cases of soft phonation, no part i a l appeared in the spectrum in
t h i s frequency region. The maximum difference was 5 dB, but mostly the
values were lower than 2 dB. The c o r r e s p d i n g values for the partial
closest t o 3 kHz are shown in Fig. Id. The data are s i m i l a r t o those
shown in Fig. lc.
These r e s u l t s indicate t h a t repeated spectrum analysis yielded
level values that mostly d i f f e r e d from each other by 1 and 2 dB, o r
less, for the lower and the higher spectrum partials, respectively.
Result s
Spectrum characteristics in loud and soft phonation
1.
Fig. 2a and 2b show two typical examples of vowel spectra from a
female subject i n soft and loud phonat ion at a fundament a1 frquency of
approximately 400 Hz.
In soft phonation, only the f i r s t two spectrum
p r t i a l s appeared a d the fundamental had the highest level. In loud
phonat ion, twelve part i a l s appeared, and it was the second part i a l which
had t h e highest level. Table A1 (see Appendix) shows, f o r t h e e n t i r e
material, the order number of the strongest spectrum partials at different fundament a1 frequencies.
PITCH
F U N D A M E N T A L F R E Q U E N C Y (Hz)
PITCH
FUNDAMENTAL FREQUENCY (Hz)
Fig. la-d. Extreme level differences observed in various frequency bands in three repeated spectrum analyses
of the same /a/ vowels as produced by two female subjects; the panels show the differences observed for the fundamental (a), for the strongest spectrum partial in loud phonation (b), for the
partial closest to 1.5 kHz (c), and for the partial closest to 3 kHz (d).
STL-QPSR 2-3/1987
F R E Q U E N C Y (kHz)
I
0
I
1
,
2
I
I
3
4
!
F R E Q U E N C Y (kHz)
Fig. 2.
Two typical examples of spectra of the vowel /a/ as
produced by a female subject in soft and loud phonation at a fundamental frequency of approximately 400 Hz.
STL-QPSR 2-3/1987
I n soft phonation, the fundamental was the loudest, almost invariably, f o r both male and female s u b j e c t s , a s expected. Toward high
fundament a1 frequencies, however, the second part i a l was occasionally
the strongest one, presumably because of a raised vocal effort ( K i t zing,
1986).
In loud phonat ion, an overtone was mostly the loudest part i a l for
both the females and the males. The order number of t h i s partial tended
t o decrease with increasing fundamental frequency as in loud phonat ion
the f i r s t formant contains the strongest spectrum partial; for a low
fundamental frequency, t h i s part i a l has a high number, and vice versa.
Females tend t o have higher formant frequencies than males because
of t h e i r shorter vocal tracts. n?us, one may expect that, for a given
fundamenta1 frequency, the st rongest part i a l w i l l have a higher number,
on t h e average, f o r women than f o r men. Fig. 3 shows t h e average
number of the strongest partial. However, the mean values were almost
identical for men a d women at a l l fundamental frequencies. T h i s sugg e s t s t h a t t h e mean formant frequency d i f f e r e n c e between t h e s e two
groups of s u b j e c t s was small. A t 131 H z fundamental frequency, t h e
This reflects the fact
females have a lower average than the males.
that i n some female subjects the fundamental was the strongest spectrum
part i a l at t h i s pitch which represented the lower l i m i t of t h e i r range.
PITCH
F U N D A M E N T A L F R E Q U E N C Y (Hz)
Fig. 3.
Average number of the strongest partial at different fundamental frequencies for ten male and ten female subjects phonating
the vowel /a/ in soft and loud phonation.
The degree t o which the overall SPL reflects the sound level of one
sinqle partial is of relevance t o a proper understanding of a phonetogram. Therefore, the level di f ference bet ween the strongest part i a l and
STL-QPSR 2-3/1987
the overall SPL was computed for a l l vocalizations in the ten male and
ten female normal voices. Fig. 4 shows the results. (hn the average, the
difference varied between O and 4 dB, depending on subject sex, pitch
and loudness. A s expected, the difference was greater in loud phonat ion
because of the greater number of strong partials. Also, i n loud phonat ion it was greater for the male subjects and, regardless of sex, it was
greater a t lower fundamental frequencies. The reason would be that a low
fundamental frequency was associated with a greater number of strong
part i a l s . I n soft phonat ion, t h e difference s l i g h t l y increased with
fundamental frequency for a l l subjects. This probably reflects a tendency t o use a higher degree of vocal effort at high pitches; t h i s would
raise the levels of overtones of the lowest spectrum, as compared with
that of the fundamental. The graph in Fig. 4 reveals that in the lower
half of the fundamental frequency range, the lower phonetogram contour
mainly reflected the voice source fundamental.
According t o acoustic theory of voice product ion, the overall SPL
increases with increasing f i r s t formant frequency (Fant , 1970). Lindblom showed that an increase in vocal effort was generally associated
with a widening of the jaw opening (quoted in Sundberg, 1970), which,
according t o measurements on an articulatory model of the vocal t r a c t ,
is of particular importance t o the frequency of the f i r s t formant (Lindblom & Sundberg, 1971). Thus, we might suspect that a t a given fundamental frequency, the f i r s t formant frequency may differ between soft
and loud phonation.
The f i r s t formant frequency was est imatd from the spectrograms of
a l l the ten male subjects' vocal izat ions at fundamental frequencies in
the vicinity of 100 Hz. In two of the subjects, the f i r s t formant could
not be clearly discerned. The results from the remaining eight subjects
are listed in Table I. It can be observed that in a l l cases the f i r s t
formant frequency was raised with increasing sound level. It can also be
seen that i n soft phonation, the f i r s t formant frequency was sometimes
as low as 400 Hz, or lower. This is far away from its normal value for
t h i s vowel, and, a s expected, t h e vowel q u a l i t y was f a r from being
typical in such cases. I n loud phonation, on the other h a d , the values
are more typical for the vowel /a/.
Fig. 5 compares the increases in SPL and f i r s t fonnant frequency,
expressed in the logarithmic semitone unit. A slight tendency can be
seen for the first formant t o increase with increasing SPL. Ebwever,
t h i s is merely a vague trend, possibly reflecting a typical behavior in
some subjects, and in some cases great SPL increases were accompanied
with a small first formant frequency change and vice versa. These formant frequency chanqes accompanying changes in vocal loudness would
complicate a n interpretat ion in phonatory terms of a phonetogram, as
they obscure t h e r e l a t i o n s h i p between t h e SPL values and t h e voice
source propert ies.
STL-QPSR 2-3/1987
0
;10
20
30-I
40
50
60
70
F U N D A M E N T A L FREQUENCY ( % o f
FUNDAMENTAL
Fig. 4.
FREQUENCY
80
90
100
t o t a l range)
( % of total range)
Average difference between SPL and the level of
the strongest partial for the vowel /a/ produced
at different fundamental frequencies in soft and
loud phonation by ten male and ten female subjects. The bars represent +/- one standard deviation.
STL-QPSR 2-3/1987
Table I.
Subject
HIP
I1
JSK
A .
II
CHANGE
Fig. 5.
IN SPL (dB)
Increases in the first formant frequency, expressed in the logarithmic semitone unit, as
function of the concomitant increases in SPL
observed when 8 male subjects phonated the
vowel /a/ in soft and loud phonation at a
fundamental frequency near 100 Hz. The line
represents the general trend, disregarding
extreme, deviant data points.
STL-QPSR 2-3/1387
Significance of using the A-weighted curve
Given the fact that the fundamental terads t o be the strongest
partial in soft phonation, we may assume that the application of an Aweighted frequency curve w i l l affect the SPL values.
The SPL values for the different vowel sounds produced by the ten
male ancl ten female subjects were determined fran the spectrum data. As
expected, the choice of a f l a t or an A-weighted frequency curve in
determining the SPL was significant t o the phonetogram curves and these
effects differed between male and female voices.
Fig. 6 shaws tw
typical examples.
With a f l a t frequency curve, the lower contour for
the female voice (Fig. 6a) was more horizontal than when the A-weighted
The difference decreased with pitch and was small in
curve was used.
the upper part of the subject's range.
For the male voice (Fig. 6b),
the SPL curve was almost horizontal aver the entire range and increased
only for the highest fundamental frequencies. For both subjects, the Aweighted c m e gave mewhat lower SPL values throughout the range in
loud phonat ion, as expected
The SPL difference, result ing fran switching fran a f l a t t o an Aweighted frequency curve in the determination of the SPI, values. can be
expected t o depend on the details of the vowel spectrum. Table AII (see
m n d i x ) lists t h i s difference at the various fundamental frequencies
for a l l male a d a l l female subjects. A great variability can be o b
served. For instance, at a fundament a1 frequency of 110 Hz, the greatest
and smallest difference for male soft phonat ion was 18 and 9 dB. Thus,
it seems risky, particularly at IONfundamental frequencies, t o convert
an SPL value for a vowel spectrum fran dB(A) t o dBl or vice versa,
simply by adding or subtracting a certain n W r of dB for each fundamental frequency, as has been done m e t imes. It is also obvious that
the influence of the frequency curve chosen varies with the vowel. This
aspect w i l l be considered in Section 3 below.
W
e have seen that the lower phonetogram contour is highly dependent
upon the level of the fundamental, which, in turn, reflects the peak
amp1itude of the flow glottogram pulses. The upper phonetogram contour,
on the other hand, is mostly determined by a spectrum overtone, which is
dependent on a different flow glottogram characteristic, viz.
the
negat ive peak amplitude of the different iated flow glottogram.
Fran a
phonatory p i n t of view, then, an interesting question is the occurrence
of cases where the applicat ion of the A-weighted curve leads t o a subst i t u t ion of the fundamental as the strongest part i a l , i .e , t o a monatory redef i n i t ion of the phonetogram.
The significance f r m t h i s point of view of using the A-weighted
curve was examined by applying t h i s curve t o the strongest spectrum
part i a l s in a l l cases, where the fundament a1 was the strongest part i a l
Then, the number of t i m e s that the applicat ion of the A-weight ed curve
changed t h i s relationship was determined. The results are plotted in
2.
.
.
i
I
.
1
STL-QPSR 2-3/1987
-
50
-
PITCH
FUNDAMENTAL FREQUENCY (Hz)
PITCH
FUNDAMENTAL FREQUENCY (Hz)
Fig. 6.
Two typical examples-of.phonetogram registrations
for a female ( a ) and a male ( b ) subject phonating
on the vowel /a/. The solid and dashed curves represent measurements obtained with a flat and an
A-weighted frequency curve.
STL-QPSR 2-3/1987
Fig. 7a - 7d. In soft phonat ion, the fundamental was the loudest part i a l almost invariably, see Fig. 3. However, applying the A-weighted
curve changes t h i s i n several instances. A s a result, one cannot take
for granted that the lower phonet ogram contour reflects the amplitude of
the source spectrum fundament a1 in soft phonat ion. I n loud phonat ion,
t h e s i g n i f i c a n c e of applying t h e A-weighted curve i s smaller. This
indicates that the A-weighted curve a l t e r s the phonatory significance of
the phonetogram.
lo
_I
+u
Z
-
a)
FEMALE. L o u o
W
a
u
0
0
z
2
iL
C
'I
3
:
01
a
1 0 7
t;
-
.
I
;
$u.
gu
t;
0
FEMALE. SOFT
0
0
10
--
D:
W
W
rJ
I
z
MALE. S O F T
*
3
I
MALE. LOUD
v
,
a
0
FUNDAMENTAL FREOUENCY (Hz)
Fig. 7.
3.
The number of spectra produced at different fundamental
frequencies by ten female and ten male subjects in which
the fundamental was the strongest partial. Solid and
dashed curves pertain to data obtained with a flat and an
A-weighted frequency curve.
Relevance of vowel
I n phonetogram measurements, d i f f e r e n t vowels a r e o f t en used.
Coleman, Mabis, & Hinson (1977) allowed the subjects t o use any vowel
sound t o produce the vocal extremes, while ,Stone & Ferch f 1982), analyzing t h e lower contour only, asked t h e s u b j e c t s t o s u s t a i n t h e vowel
[i:], and Seidner, Kriiger, & Wernecke (1985) used the vowels [a:], [u:),
and [i: 1. I n most cases, however, the vowel [a:] has been preferred
amst st& 1-970: Heinemann & Gabriel, 1982; Komiyama, Watanahe, & Ryu,
1984; Ohlsson & Wfqvist, 1986).
It has been recommended that usinq the vowels /a, i, u/ for phonetogram measurement s a r e very informat ive (Schutt e & Seidner, 1983).
Thus, it is relevant t o analyze somewhat i n detail what the effects are
of the choice of vowel on the phonetogram contours.
STL-QPSR 2-3/1987
If documentation of the functioning of the voice source is the goal
for recording a phonetogram, i t is not clear t o what extent an additiona l vowel w i l l add any informat ion; the voice source would be basically
similar for different vowels, and the SPL difference between the vowels
would depend mainly on the differences in the formant frequencies bet ween the vowels. Interact ion effects between the vocal tract and the
voice source do occur (see, e.g., Titze, 1985) but would be of s e c o d a q
interest in most applicat ions of phonetogram measurements.
According t o Fant , F i n t o f t , L i l jencrant s , Lindblom, & M&rtony
(1983), the first formant as function of fundamental frequency reaches
peaks at frequencies where a spectrum partial equals the f i r s t formant
frequency and troughs when the formant is midway between two adjacent
part ials. ?he level difference bet ween such peaks and troughs may amount
t o no l e s s than 14 dR. As, mostly, t h e strongest p a r t i a l i n a vowel
spectrum is that lying closest t o the f i r s t formant, and t h i s part id
generally almost determines the overall SF%, the level variation of the
f i r s t formant caused by the spectrum details is relevant t o the phonetogram contours. Such level variat ions were previously demonstrated by
means of a model experiment carried out on a formant synthesizer (Gramming, Gauffin, & Sundberg, 1986). For soft phonation at low and moderat e l y high fundamental frequencies, t h i s effect would be less clear or
even missing, as the fundamental is often the strongest spectrum partial
in t h i s case.
Fig. 8 illustrates the relevance of the vowel t o the phonetogram
contours in terms of averages for 22 speech t h e r a p i s t students. The
average SPL, computed over fundamenta1 frequency and vowel, is sl igthly
lower than the averages over fundamental frequency reported for these
vowels by Seidner & a l . (1985). The reason f o r t h i s difference i s
probably t h a t t h e Seidner averages disresarded t h e subjects' lowest
fundamental frequencies. The figure also shows that the vowel /a/ gives
higher SPL values than the other vowels used. The difference is about 10
dB a t low fundament a1 f rguencies and decreases towards higher f undamental frequencies. T h i s is in qualitative agreement with the effect of
the frequency of the f i r s t formant on the SPL.
However, these phonet ograms were recorded using the ~ R ( A )weighted
frequency curve which af fect s those vowels most, which have the lowest
f i r s t formant frequency. Therefore, part of t h i s effect is due t o the Aweight d curve.
A simple experiment was made i n order t o f u r t h e r elucidate t h e
effect of the formant frequencies on the contours. Usinq a linear frequency curve, phonetoqrams were recorded for two healthy subjects, one
male and one female, using the vowels /a, i, e, u/.
Subsequently, a
spectrum analysis was made of a l l the vocalizations by means of the B&K
2033 spectrum analyzer described above using both t h e dB(A) and t h e
linear frequency weighted curve.
STL-QPSR 2-3/1987
-
53
F U N D A M E N T A L FREQUENCY
-
( % of t o t a l r a n g e )
Fig. 8. Averaged phonetograms for the vowels /a/ (boxes), /u/
(circles), and /i/ (triangles) produced by 22 female
speech therapy students. The data were obtained using
an A-weighted frequency curve. The dashed and dotted
lines represent the average over fundamental frequency
and vowel for the present material and for data reported by Seidner & al. (1985).
Fis. 9 i l l u s t r a t e s the effect on the different vowels of using the
two frequency curves for the female subject. In the case of the /a/,
t h e e f f e c t i s smaller than i n t h e t h r e e other vowels. This is i n
accordance with the expectations. I n an /a/, the high f i r s t formant
frequency yields spectra with a high frequency of the strongest spectrum
p a r t i a l and at high frequencies, the difference between the two frguency curves is small. Hence, a small level difference results.
In the
other vowels, the strongest partial occurs at a lower frequency and the
difference becomes greater.
Fig. 10 compares the phonetograms recorded with the linear frequency curve. As SPL increases with the frequency of the f i r s t formant,
other things being equal, we would expect the vowels /u, e, i / , having
low f i r s t formant frequencies, t o produce lower SPL values than the /a/.
This effect can be observed in loud phonation but i s missing in soft
phonation, as, in t h i s case, the f i r s t formant does not affect the SPL.
At the highest fundamental frequencies, t h i s does not apply. In
these cases the relat ionship bet ween the two lowest formant frequencies
and the frequencies of the spectrum partials seems t o offer an explanation. The low SPL values in the /a/ produced by the female subject was
probably caused by the fact that there was no partial close t o the f i r s t
formant, so the strongest partial f e l l in the second formant. For the
other vowels i n t h i s pitch ranqe, the fundamental was the stronqest
partial. Thus, these differences can be explained by t h e spectrum
details.
I
I
I
I
I
1
I
I
STL-QPSR 2-3/1987
For t h e female subject, t h e upper contour e x h i b i t s peaks and
troughs at certain fundament a1 frequencies. Such discont inuit ies are
somet imes regarded as s iqns of register transit ions (Klingholz, Mart in,
& ,Jolk, 1985). However, in t h i s case, these dips i n the upper contour
apparently depended on the spectrum details. In most of cases of contour
discontinuities, the peaks were associated with spectra with one single
very dominating partial, thus suggesting that t h i s partial matched the
frequency of the formant. Troughs were associated with spectra i n which
the level difference between the two s t r o v e s t spectrum p a r t i a l s was
small, suggesting that these partials f e l l symmetrically on the s k i r t s
of the formant.
In soft phonat ion at high fundament a1 frequencies, the differences
become considerable between the vowels. A general observation was that
high SPL values were associated with spectra with one very strong part i a l , again suggest ing that a partial f e l l in the center of a formant,
while low SPL values were associated with spectra where there were two
or more partials of equal or almost equal strength. Ebwever, there were
several cases which did not adhere t o t h i s pattern.
Discussion
The choice of vowel is by no means insignificant t o the results.
First, according t o classical theory of voice production, a one-octave
r i s e of the f i r s t formant frequency leads t o an SPL increase of 3 dBI
other things being equal. Also, we have seen that the significance of
using a f l a t o r an A-weighted frequency curve depends on the vowel.
A possible advantaqe of using t h e vowel [a:] might be t h e high
frequency of the f i r s t formant, approximately 700 Hz. T h i s means that,
at least for male voices, the fundamental frequency is automatically
lower than the f i r s t formant almost throughout the range, and the relation between SPI, and the voice source is reasonably simple.
A vowel having a low frequency of the f i r s t formant, such as [i:]
or [u:] , may entail certain complications. Its SPL w i l l automatically
drop i f t h e frequency of t h e fundamental i s higher than t h a t of t h e
f i r s t formant. 01 the other hand, the subject may raise the f i r s t formant in such cases so that it always remains higher than the fundamenta l , a strategy typically used by soprano singers (Sundberg, 1975). I n
such cases, the SPL w i l l increase considerably without any increase of
vocal effort. Thus, i n vowels with a low f i r s t formant frequency produced at high fundamental frequencies, the relationship between SPL and
the vocal effort may be obscured by the individual subject! s a r t iculatory habits. This w i l l complicate an interpretat ion i n phonatory t e m s
of the phonetogram.
Another advantage with chosinq a vowel with a high f i r s t formant
frequency may be that the upper and lower phonetogram contours more
clearly reflect different aspects of the voice source; if a vowel with a
STL-QPSR 2-3/1987
high f i r s t formant is used, the upper contour is likely t o reflect an
overtone even at rather high fundament a1 frequencies, while the lower
emtour mainly depends on the Fundament al.
W
e have seen that the significance of using an A-weighted rather
than a f l a t frequency curve i n determining the SPL is great. tbwever,
the effcct is by no means predictable, as it depends on the frequencies
of both the fiindamental and the f i r s t formant. If in loud phonation the
f i r s t formant frequency i s just midway bet ween two part ials, these part i a l s w i l l be the stronqest ores in the spectrum. Also, they would be
similar in amplitude, but l e t us assume that the lower one is s l i a h t l y
stronger. The A-weight ed curve imp1ies that at t enuat ion decrease with
increasing frequency, so that the lower partial w i l l lose more i n level
than the higher one. Tnis difference i n level reduction w i l l tend t o
make the higher part i a l the loudest. This case was found t o be common.
In some cases, on t h e other hand, t h e r e was no difference, probably
because one s i n g l e p a r t i a l was very c l o s e t o t h e f i r s t formant f r e quency.
An advactage with usina the ciB(A) is a reduct ion of the influence
from low frequency ambient noise in the recording room. tbwever, it
miqht be preferable t o redtlce the amust i c informat ion during the analysis r-ther than already at the recording. In particular, t h i s is true
if a soundproof room is available, where the ambient noise is minimized.
The use of the A-weighted curve has a stronger impact on the lower
rontour of a phonetogram. This must be observed when comparing phonetograms. Also, by using dB(A), the slope of the lower contour gets steepe r so that the distance t o the upper contourg and, hence, the voice area
("St immfeld") appears t o be qreater.
A further disadvafitage with using the A-weighted frequency curve is
that the phonetgraq contours reflect larynegeal charact e r i s t ics of a
voice in a more complex way than when a linear frequency curve is used.
Summarizing, it seems t h a t t h e use of t h e A-weighted frequency
curve entails severe d i f f i c u l t i e s which miqht inflict on the clincal
usefulness of IJnonetoqrams.
Above we have discussed f a c t o r s a f f e c t inu t h e p o s s i h i l i t i e s of
deriving phonator' infomat ion from a phonet ogram, assuming t a c i t l y that
t h i s is the u l t imate goal of phonetography. It can he argued, however,
t h a t t h i s is not t h e only r a t ionale f o r phonetogram recording. The
sheer p o s s i b i l i t y of a voice t c produce loud sounds a r e per se of
relevance from a communication p i n t of view, reqardless of the underlyinq type of voice use. On the other hand, the way the voice organ is
used, i~ rt icula rl y i n logd phonat ion, wouli! represent a major concern in
phoniat r ic pract ice.
STL-QPSR 2-3/1987
Conclusions
The two phonetogram contours, representing soft and loud phonation,
may inform about t w o d i f f e r e n t a s p e c t s of t h e g l o t t a l v o i c e source. For
t h e lower contour, t h e SPL is mainly dependent on t h e amplitude of the
fundamental which is r e l a t e d t o t h e peak amplitude of t h e flow g l o t t o gram. F o r the u p p e r c o n t o u r , t h e SPL is m a i n l y d e t e r m i n e d b y v o i c e
source spectrum overtones which depend on the negative peak amplitude o f
t h e d i f f e r e n t iat ed flow glottoqram.
These r e l a t i o n s h i p s a r e strengthened i f a vowel is used which h a s a
h i g h f i r s t formant frequency, and i f t h e SPL is measured using a f l a t
frequency curve r a t h e r t h a n a n A-weighted curve. Cn t h e o t h e r hand, t h e
r e l a t i o n s h i p s are complicated by t h e f a c t that t h e f i r s t formant frequency t e n d s t o increase w i t h increasing loudness o f phonation.
Peaks and troughs i n a phonetogram contour cannot always be i n t e r p r e t e d as a s i g n of r e g i s t e r s h i f t s . They s o m e t i m e s occur f o r p u r e l y
a c o u s t i c a l r e a s o n s d e p e n d i n g on t h e f r e q u e n c y d i s t a n c e bet ween the
s t r o n g e s t p a r t i a l s and the closest formant frequency.
Acknuwledgment s
W e g r a t e f u l l y acknowledge t h e Department of Fboniat r ics , M a l m o
General I-bspit a1 f o r providing access t o recording and computer f a c i l i ties, i n p a r t i c u l a r Lab.eng. I;. Akerlund f o r t e c h n i c a l a s s i s t a n c e i n
measurement and r e c o r d i n g , and P r o f e s s o r N.G. Toremalm f o r v a l u a b l e
s u p p r t i n many d i f f e r e n t ways.
References
Coleman, R.F., Mabis, J.H., & Hinson, J.K. (1977). "Fundamental f r e voices,"
quency-sound p r e s s u r e l e v e l p r o f i l e s o f a d u l t m a l e and female
J.Speech Hear.Res. 20, pp. 197-204.
D a m s t &, P.H.
pp. 185-187.
(1970). "The Phonetogram," P r a c t ica Oto-Rhino-Lar.
32,
Fant , G. (1970). Acoust ic Theory o f Speech P r o d u c t i o n , (Mouton, The
Hague) (2nd e d i t ion).
F a n t , G., F i n t o f t , K., L i l j e n c r a n t s , J., Lindblom, B.,
(1963). " ~ o r m a n t amp1it ude measurement s," J.Acoust .Soc.Arn.
1761.
& p4&-tony,
35,
J.
pp. i753-
Fant, G., L i l jencrants, J., & Lin, 0. (1985). "A four-parameter model o f
g l o t t a l flow," STLrQPSR 4/1985, pp. 1-13.
G a u f f i n , ,J. & Sundberg, J. (1980). "Data on the g l o t t a l v o i c e s o u r c e
behavior i n vowel product ion, " S W I ? S R 2-3/1980, pp. 60-70.
Gramming, P., Gauffin, J., & Sundberq, J. (1986). "An attempt to improve
14, pp. 421t h e c l i n i c a l usefulness of phonetograms," J. of Phonetics 427.
STL-QPSR 2-3/1987
Heinemann, M. & G a b r i e l , H. (1982). "Moglichkeiten und Grenzen d e r
St immfeldmessung-Vorst ellung des Ileiserkeit s f e l d e s a l s Erg%~zung der
Methode," Sprache-St imme-Geht5r 6, pp. 37-42.
Kitzing, P. (1986). "LTL5 c r i t e r i a pertinent t o t h e measurement of voice
14, pp. 477-482.
quality," J. of Phonetics Klingholz, F. & Mart i n , F. (1983). "Die q u a n t i t a t i v e Auswertung d e r
7, pp. 106-110.
St immfeldmessung," Sprache-St imme-Gehor Klingholz, F., Mart in, F., & Jolk, A (1985). "Die Best immung d e r Regist erbruche aus dem St immfeld," Sprache-St imme-Gehor 9, pp. 109-111.
Komiyama, S., Watanabe, HeI & Ryu, S. (1984). "Phonetographic relation36,
s h i p between pitch and i n t e n s i t y of t h e human voice," Folia Phoniat. pp. 1-7.
Lindblom, B.E.F. & Sundberg, ,J. (1971). "Amust i c a l consequences of 1ip,
tonque, jaw, and larynx movement," J.Acoust. Soc.Arn. 50, pp. 1166-1179.
Chlsson, A-C. & m f q v i s t , A. (1986). "Phonetograms of normal and pathological voices," Working papers i n Lqopedics and phoniatrics 3 (Univers i t y of Lund), pp. 94-106.
Schutte, H.K. & Seidner, W. (1983). "Recommendat ion by t h e Union
of
European Phoniat r i c i a n s (UEP): Standardizing Voice Area Measurement /
Phonetography, " Fol.Phoniat
35, pp. 286-288.
.
Seidner, W., Kriiger, H e , & Wernecke, K-D. (1985). "Numerisdhe Auswertung
spekt r a l e r St immfelder, " Sprache-St imme-Gehor 9, pp. 10-13.
Stone, RE. & Ferch, P.A.K.
(1982). "Intra-subject v a r i a b i l i t y i n W S P L
min voice profilesIVJ.Speech Hear.Dis. 47, pp. 134-137.
SuncXberg, J. (1970). "Formant s t r u c t u r e and a r t i c u l a t i o n i n spoken
22, pp. 28-48.
sung vowels," Fol.Phoniat. -
and
Sundberq, J. (1975): "Formant t e c h n i q u e i n a p r o f e s s i o n a l f e m a l e
singer", Acust i c a 32, pp. 89-96.
Titze, I. (1985). "The i m p r t a n c e of vocal t r a c t loading i n maintaining
v o c a l f o l d o s c i l l a t i o n " , pp.
- - 61-72 i n (A. Askenfelt, S. F e l i c e t t i , E.
Jansson, & J. Sundhrq, eds.): Roc. of t h e Stockholm Elusic Arx>ustics
Conference 1983 (SMAC 8 3 ) , Publ. No. 46:l issued by t h e Edoyal Swedish
Academy of Eus i c , Stockholm.
STL-QPSR 2-3/1987
APPENDIX
.
Table A I
The order n W e r of t h e strongest spectrum part ial i n t h e d i f f e r e n t v-1
sounds at t h e fundament a1 f r e quencies specified.
Frequency: 66 8 3 9 8 110 1 3 1 165 196 220 262 330 392 440 524 660 784 880 1048
MATE SUBTECTS, LOUD PHONATION:
A.Rf4
-- 8 7 7 6 6 4 4 3 3 2-----------------------8 4 6 2 , 2 4 3 2 2 2 ----------------RMij
cJGR
11 9 7 ?t
5
6
4
5
4
2
2
3
1 I------------JCH
--- 7 6 5 4 & 8 4 3 3 2 2 2 2 ------------------------1&5 5 4 3 3
HIP
KAK
--- 7 7 6 5 4 3 ......................................
5
5
4
4
3
2
2
...........................
BaK
----5
5
4
4&3
3
3
3
3
2
1 2
1 ------------JSK
----7
6
5
4
3
3
3
2
2
1
CHA
-----8&117 6 4 4 3 3 2 2 ----------------------I&
----------------------------me---------
--------
-------------------
SOFT PHONATION:
ARA
l3r@
JGR
JCH
HIP
Kik
B'Q
JSK
CHA
1A.K
---
1
1
1
1
1 1 1 1
-- 1 1 1
1 1
1
1 1
------ 1 1
----- 1 1
-------- 1
----- 1 1
--------
I
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
FEMALE SUBJECTS, LOUD PHONATION:
WA
IIN
mE
WGi
YNI
m1
ABE
BBL
RSA
A
m
-----------------------
1
1
-----------
5
1
-----------------------------
-----------
----------- 4
---------------
SOFT PHONATION:
M A
IIN
KRE
m
YNI
FILI
Am
BRL
RSA
ACL
4
3
1
3
1
5
4&33
4
3
4 4 6
4
3
4
3
4
3
6
3
......................
1 1 -------------1 1 I------------2
-----------------.......................................
......................................
2
2
2
...........................
1 1 2
2
1 1 1 ------------1 1 1 1 1 -----------------1 1 1 1 ......................
1
1
1
1
1
1
1
2
1
1
1
2
1
1
1
2
3
3
2
2
......................
3
3
2
3
2
2 ---------------3&2 4
2
2
2
2
1 ----------3
2
2
2
2
2
-------------3
3
2
2 1&2 1 1
& 3 3 3
2
2
2
2
2
2--3
3
2
3
2
2
1
1
1
1
3
3
3
2
1
2
2
1
1
1
3
2
2
2
2
2
1
1
1
3
2
2
2
1
2
1
1
1
-
----------
STL-QPSR 2-3/1987
Table AII. Difference in SPL values caused by shifting fran a flat to an
A-weighted frequency curve.
66 83 98 110 1 3 1 165 196 220 262 330 392 440 524 660 784 880 1048
IIALE: SUBTEGS, LOUD PHONATION
AF&
-- 2 0 0 1 0 1 1 1 0 0 .......................
~ 9 3
--------- 1 3 2 3 2 1 0 1 0 0 ----------------JGR
2
0 1 2
1 0
1 0
1 1 1 1 1 2 ------------3 3 4 1 & 8 1
1
2
2
1 1 1 ------------------JCH
HIP
------ 6 5 3 2 2 .......................................
--- 3 2 2 2 2 2
KAI<
-------4
2
1
2
1
1
1
...........................
BaK
---5
3
4
3
2
1
2
1
2
3
1 2------------JSK
----2
2
2
4
2
2
2
1
1
2
CHA
----- 0 0 0 2 2 1 1 1 2
LAX
F'requency
---
--------------------me-----------------
------------------
---------------------em
SOFT PHONATION
A
m
l3m
JGR
JCH
HIP
KPX
BaK
JSK
CHA
LftK
---
20 17 1 4
17
2 1 20 20 1 6
1 8 17 17
----- 1 5 17
-- 7 13 12
----- 9 9
----- 1 8 16
-------- 1 4
----- 19 18
16
16
16
12
11
15
6
13
14
16
--------
--
11 9
12 11
13 6
8 8
11 7
8 8
9 4
11 8
12 11
14 9
Frequency 66 83 98 110 1 3 1 165 196 220 262 330 392 440 524 660 784 880 1048
FEMF-I;E , S U B ~ S ,LOUD PHONATION
------------11 2 1 1 1 I s - - - - RJA
------------ 1 1 6 4 2 2 3 0 1 0
IIN
---------me-----
KRE
ma
YNI
m.81
ABE
BBL
RSA
ACT,
-------------------------------------------
-----------
1
---------------
------------- 2
---------------
IIN
KRE
m&
YNI
MIJI
ARE
RBL
RTA
AQ;
------------
------------
5
15
15
--------------------------------------------------
-----------
13
15
------------------------- 1 4
---------------
1
2
2
1
---------------1
SOFT PHONATION
WA
1
5
2
1
1
1
1
1
1
1
2
1
3
4
2
1
0
2
1
2
2
1
1
1
1
2
1
S
O
1
1
1
2
2
0
2
3
1
1
2
1
1
1
1 1
1
0
0
1
2
1
1
0
2
3
1
0
1
2
1
0
1
1
1
1 ------------------------2 -----------1
0
0---0
0
0
0
0
0
0
0
1
0
0
1
-