Beruflich Dokumente
Kultur Dokumente
NAOAKI YANAGIHARA
ANALYSIS OF HOARSENESS
Method
The hoarse voice samples were recorded on magnetic tape using a dynamic
microphone (FP-1, Sony). The voice was recorded at a distance of 20 cm from
the microphone.
531
Five vowels/u, o, a, e, i/were selected. The subject was instructed to pho-
hate the five vowels: (1) gliding from /u/ to /i/ and (~) sustaining each
vowel for several seconds. These utterances were carried out in the most com-
fortable manner with no restriction imposed on pitch or loudness.
The recordings of the entire series of vowels produced by 167 patients with
hoarseness were subjected to perceptual evaluation of the degree of hoarse-
ness. The judges were three otolaryngologists with ample clinical experience.
They rated each subject as having a slight, moderate, or severe degree of
hoarseness.
From the total number of 167 cases, 10 with each degree of hoarseness-
slight, moderate, and severe-were selected with the following two precautions:
(1) In each case, all three judges agreed on the degree of hoarseness, (2) Sex
and age distribution were as similar as possible among the three groups. The
selected cases included 16 male and 14 female subjects ranging in age from
18 to 60 years. Sonagrams of the entire series of vowels were made under
constant conditions using a narrow band filter (45 Hz). Sections were taken
at the approximate midpoint of the vowels/u/,/a/, and/e/. The fundamental
frequency of voice was measured at these points. The average value of the
fundamental frequencies of these three vowels was taken to be the fundamental
frequency of comfortable phonation.
Results
Figure 1 illustrates the sonagrams of the sustained v o w e l s / u / , / a / , and lel
phonated by a normal female subject at a fundamental frequency of 250 Hz.
Compared with these normal patterns, the sonagrams of the v o w e l s / u / , / a / ,
a n d / e l phonated by a very hoarse female patient at fundamental frequency
of 210 Hz (Figure 2) indicate three distinctive features: (1) There is strong
noise energy above 5000 Hz in the vowels /a/ and /e/, (2) The harmonic
structures between 1000 Hz and 3000 Hz i n / a / (2nd and 3rd formants) and
between 2000 Hz and 3000 Hz in /e/ (2nd and 3rd formants) are almost
completely replaced by noise components, and (3) In the pattern of the
vowel/u/, there is scarcely any structural change in the harmonic components
and no remarkable noise component.
Throughout the observation of these 30 cases, the noise components represent
pertinent findings of hoarseness. The range and energy of the noise eom-
portents vary with the perceptual degree of hoarseness, and they are more
evident in the v o w e l s / a / , /e/, a n d / i / than in the v o w e l s / u l a n d / ~ / .
In slight hoarseness the noise components are limited to the main formant
ranges in these five vowels, especially to the second and third formants, and
there is no remarkable noise component in the high frequency range. Figure
3 illustrates these findings by a representative sonagram. With the progression
I
~ and intensified noise components, and loss
of high frequency harmonic structure are
i characteristic features. The pitch of voice is
approximately 210 I-Iz (a). Clinical diagnosis
~'c is eordal cancer, and acoustic diagnosis is
hoarseness Type IV.
I~ SEC
Spectrographic Perceptual F. F.
No. Type Degree Age Sex in Hz ClinicalDiagnosis
1 Slight 18 F 214 Acute Laryngitis
2 Slight 55 M 143 Chronic Laryngitis
3 Type I Slight 34 F 214 Acute Laryngitis
4 Slight 23 F 286 Acute Laryngitis
5 Slight 54 M 150 Papilloma of Larynx
6 Slight 46 M 143 Vocal Nodules
7 Slight 37 F 286 Paralysis of Vocal Cord
8 Slight 60 M 164 Vocal Cord Cancer
9 Slight 49 F 178 Chronic Laryngitis
I0 Slight 29 M 143 Vocal Nodules
11 Moderate 31 F 178 Chronic Laryngitis
12 Type II Moderate 36 M 143 Chronic Laryngitis
13 Moderate 49 M 128 Chronic Laryngitis
14 Moderate 37 F 286 Paralysis of Vocal Cord
15 Moderate 39 M 150 Paralysis of Vocalis Muscle
16 Moderate 63 F 200 Vocal Cord Cancer
17 Moderate 39 M 107 Hyperplastic Laryngitis
18 Moderate 31 F 214 Hyperplastic Laryngitis
19 Moderate 64 M 143 Vocal Cord Cancer
20 Moderate 36 M 118 Paralysis of Vocal Cord
21 Type III Severe 34 F 286 Vocal Nodules
22 Severe 51 M 143 Hyperplastic Laryngitis
23 Severe 37 F 214 Paralysis of Vocal Cord
24 Severe 40 F 178 Vocal Cord Polyp
25 Severe 48 F 214 Vocal Cord Cancer
26 Severe 40 M 143 Paralysis of Vocal Cord
27 Type IV Severe 62 M * Vocal Cord Cancer
28 Severe 59 M * Atrophic Laryngitis
29 Severe 55 M 107 Paralysis of Vocal Cord
30 Severe 45 F 143 Hyperplastic Laryngitis
" u o a ,
SYNTHESIS OF HOARSENESS
Method
The significance of additional noise components and changes in the harmonic
structure revealed by the sound spectrographic analysis was supplemented by
a synthetic study. An outline of the method is illustrated in Figure 6. The
synthesized hoarseness was prepared from a normal male voice and band-pass
filtered noise, recorded at constant intensity level. As shown by Figure 6, a
tape recorder (A) was used for the reproduction of the normal vowels /u/,
/a/, a n d / e / , and a similar tape recorder (B) for the reproduction of the full
band noise. Filtered noise and nonfiltered or low pass filtered vowels were
recorded on each channel of the endless loop tape and reproduced by a two-
Type I
F, F~ Fa High ErequencyNoise
u 200 ( ~ 700Hz
a 300 c .... ~ 1700
e 200<--->700 1600 ( ) 2600
Type H
F~ F~ F~ High Frequency No~e
u 200 ( ) 800Hz
n 300 c > 1500 2800<-->3700
e 200<-->600 1800 ( > 2800 3100<-->4100 5700<--->6300
Type III
F, F~ F:, High Frequency Noise
u 200 ( ) 900Hz
a 400 ( ~ 1600 2400<---~3500 4400<--~4800 5500 ( )
e 200<---~900 2000 ( ) 3000 3500(---)4300 5200 ( )
Type IV
F, F._, F., High Frequency Noise
u 200 ( ) 1200Hz
(~ 400 ( ) 1600 2000(---)3600 3900(---)4600 5600 ( )
200~--~900 1400 ( "~ 3000 3200(---)4400 5100 ( )
A R / A N t t m X l t o ~ s v N t ~ t s l s OF H o * n s t ~ ( $ S
rutt r~Act N - r~
I~'R0~ICIIOI~ ~ FUlL10t~ PI01S[ ~
~ t m t t smrMtsltt0 ra~t
ludge 2 3 4 5 6
No. 1 0.81 0.82 0.84 0.79 0.86
2 0.81 0.63 0.45 0.73
3 0.73 0.94 0.84
4 0.93 0.79
5 0.53
Average interjudge agreement = 0.77
channel tape recorder. The output of each channel of this tape recorder was
fed to another tape recorder (C) through the mixer. Before mixing, the in-
tensity level of band-filtered noise was attenuated to a predetermined value
Results
Interjudge agreement on the hoarseness rating was tested by calculating the
reliability coeffcient (McNemar, 1962, pp. 145-148). The coefficient was ob-
tained by correlating each judge's rating for each tone with the rating by the
five other judges. These correlations are given in Table 3. The average eoeff-
cient of 0.77 demonstrated that the interjudge agreement was close enough to
permit the use of averaged rating scores as indicators of the listener's responses.
Sound spectrograms of all the test tones were made to see the result of the
synthesis and to correlate the acoustic structure of each test tone with the
corresponding judgment of the listeners.
Tables 4, 5, and 6 show the results of this experiment. Any test tones which
were checked by any one of the listeners as "artificial" were excluded from
these tables.
The significance of the noise components in the formant ranges is shown in
'Fable 4. When the band-filtered noise intrudes on the formant range, the sound
is perceived as a very slight or slight hoarseness. In the case of v o w e l s / a / a n d
/e/, mixing of the band-filtered noise into the second formant region results
in an increase of the degree of hoarseness.
In Table 5, the effect of elimination of high frequency harmonics of vowels
on the perceptual hoarseness is shown. Even ff the relative intensity of the
noise components and the harmonic components remain unchanged, the loss
of high frequency harmonics results in an increase of the degree of perceived
hoarseness. When the elimination includes the second formants of these vowels,
an advance in the degree of perceived hoarseness results.
In the case of severe hoarseness, an expansion of the additional noise com-
ponent is an outstanding feature. In order to reproduce this feature, band-pass
noise and low-pass noise were superimposed on the low-pass vowel-band
filtered noise mixtures. Table 6 shows the effect of this procedure. All test
tones of this group are perceived as either very hoarse or extremely hoarse.
Figures 7 and 8 illustrate the sonagrams and sections of synthesized hoarse-
ness of the vowel /e/. Their schematic drawing is given in Figure 9, which
shows step-by-step increase of the perceptual degree of hoarseness and the
corresponding acoustical structure changes.
2 F9 .............
I
o
FICVaE 9. Schematic drawing of the
sonagrams of synthesized hoarseness. Step by
kr MEAN OF RATING MEAN OF RATING MEAN OF RATING step rise of the degree of perceptual hoarse-
," 2 5 o, 2 8
4.4
or ness and corresponding acoustical structure
changes of vowel sound are shown.
t .................. - _ ~ " " :-:'~"
o ,t,,,===,===~ ................. ~
DISCUSSION
The major acoustic factors relating to hoarseness consist of (1) noise com-
ponents, and (2) loss of harmonic components. As to the noise components,
the intensity and range differ from vowel to vowel and also vary with the
degree of hoarseness. Generally, the noise element is much more evident in the
vowels / a / , /e/, a n d / i / than in the vowels [u/ a n d / ~ / . The present study
reveals that the noise components appear in the formant ranges, especially in
the second and third formant ranges at the initial stage of hoarseness. With
the progression of hoarseness, the noise components predominate the harmonic
This research was supported by Public Health Service Research Grant No. NB 04430-04
from the National Institute of NeurologicDiseases and Blindness.
REFERENCES
~ , E., and SCHLOSSX-L~UE~,B., Unregelm~issigeStinamlippenschwingungenbei Funk-
tionellen Stimmst~rungen.Z. Laryng. Rhinol., 40, 919-934 (1961).
FLANAC,',~r,J. L., Someproperties of the glottal sound source. 1. Speech Hearing Res., 1, 99-
116 (1958).
F~Nx, D. I., Hoarseness-a new classificationand a brief report of four interesting cases.
La~ngoscope, 50, 472-478 (1940).
Naoaki Yanagihara
J Speech Hear Res 1967;10;531-541