e port resume - eric · r e port resume. ed 011 011. studies on the speech of the deaf. by-...
TRANSCRIPT
R E PORT RESUMEED 011 011STUDIES ON THE SPEECH OF THE DEAF.BY- MARTONY, J.
EDRS PRICE MF$0.25 HC -$1.04 24P.
AL 000 077
PUB DATE JAN 66
DESCRIPTORS.... *SPECTROGRAMS, *PHONEMES, SPEECH, *ARTICULATION(SPEECH), *DEAF RESEARCH, *ACOUSTIC PHONETICS, SPEECHHANDICAPS, ROYAL INSTITUTE, SONAGRAMS
A COMPARISON OF SEGMENT FEATURES IN THE SPEECH CHAIN OF'WEE DEAF -BORN BOYS WITH THOSE OF THREE NORMAL- HEARING BOYSREVEALS THAT THE DEAF -BORN HAVE SPEECH PROBLEMS ASSOCIATEDWITH A LACK OF SYNCHRONY BETWEEN ARTICULATION AND PHONATION.IN ORDER TO DETERMINE THE DIFFERENCES BETWEEN THE TWO GROUPS(BOTH REPEATING THE SAME SWEDISH SENTENCES), A SPECTROGRAMSEGMENTATION METHOD WAS USED, ANALYZING THE MANNER OFPRODUCTION AND CORRESPONDING ACOUSTIC CORRELATES OF THESOURCE FEATURES AND RESONATOR FEATURES. SPECTROGRAMS REVEALEDTHAT THE DEAF BOYS' SPEECH DEVIATED FROM NORMAL PATTERNS INVARIOUS WAYS, PARTICULARLY IN AN ERRATIC TRANSITION IN SOMEPARTS OF THE VOWEL. IN ADDITION, THE ARTICULATION WAS NOTALWAYS UNDER CONTROL, AND THE DEAF BOYS HAD DIFFICULTY WITHFRICATIVE ASPIRATION. THE SECOND PART OF THIS TWO -PART STUDYDESCRIBES A SPECTROGRAM SEGMENTATION TECHNIQUE IN WHICHSEGMENT BOUNDARIES COINCIDE WITH A CHANGE IN THE FEATURECOMPOSITION. AN EARLIER MODEL WAS LIMITED BY THE FACT THATSLIGHT SHIFTS IN THE TIMING OF FEATURES CAUSED NEW SEGMENTSTO ARISE. ALSO, TkME DURATION OF SEGMENTS WAS NOT INDICATED.THE REPRESENTATION OF FEATURES AS A FUNCTION OF TIME WASCONSIDERED MORE ADVANTAGEOUS, AND THEREFORE, A COMPARISON OFTHE DURATION OF PHONEMES BETWEEN NORMAL HEARING AND DEAFSUBJECTS WAS CARRIED OUT. THE RESULTS SHOW A MUCH LOWER RATEOF SPEAKING FOR DEAF SUBJECTS. THIS DOCUMENT WAS PUBLISHED INQUARTERLY PROGRESS AND STATUS REPORTS, OCTOBER 1965 ANDJANUARY 1966, BY THE SPEECH TRANSMISSION LABORATORY, ROYALINSTITUTE OF TECHNOLOGY, STOCKHOLM. (FS)
r-2/V (2d44'7 6. RA. y PW4.6;s4E-6.5 //No 67-4 TVs 1fPoe.7--- Oct 44/417 510.6-Eem, 77'/PAISMI/SIS74,41 4/98e/Pp9roRy, /r, x i2 2-4/57/72/7-4- 7licfriA/446:y si4eiv*0
C. STUDE3S ON THE SPEECH OF YHE NAP
J. Mirtony
J..11v2:0aUCtion
The speech of deaf-born poeIge, differs in manur 'lays
from the speech of normal-hearing subjects. The following stt,:.dy
is an attempt to specify this deviation on an objective level.
For the description of the speech material a spectrogram seg-
mentation method has been used, and the segment features in the
utterance of deaf and hearing subjects have been compared. The
study is of introductory nature only.
The acoustic chain of speech contains successive min-
imal speech units (segments) Nith different characteristics
(features). By definition the segment boundaries coincide with
a change in the feature composition. The features describe the
relevant qualities of the acoustic signal which cai be corre-
lated with the speech riroduetion as well as with the speech
perception. This method of segmentation has been wor&ed out by
.pant and Lindblom (1).
There are two typos of features used for description
of the segments:
1. Segment type features which describe segments by
their manner of production, respectively the
corresponding acoustic correlates.
2. Segment prittern features related to the place ofarticulation, and specified by their acousticcorrelates.
This study is concerned with ,the first category only,
i.e. segment type features. The description of the segments in
a speech chain in tee of those features contains no information
about the prosodic features, such as stress, intonation, and
rhythm.
The segment typo features are the following:
Source features:U.S. DEPARTMENT OF HEALTH, EDUCATION it WELFARE
OFFICE OF EDUCATION
THIS DOCUMENT HAS BEEN REPRODUCED EXACTLY AS RECEIVED FROM THE
PERSON OR ORGANIZATION ORIGINATING IT. POINTS OF VIEW OR OPINIONS
STATED DO NOT NECESSARILY REPRESENT OFFICIAL OFFICE OF EDUCATION
POSITION OR POLICY. VIL 000 0 77-
1. voice
2. noise
3. transient
Resonator features: occlusive
5. fricative
6. lateral
7. nasal
8. vowel-like
9. transitional
17.
The material used in this study is the second sentence
of the text: "Dot ar vackert vier idag. Solon skiner, men dot ar
kallt".
This text has been recorded, spoken by three boys with
nnxmal hearing (111, H2, H3) and by three deaf boys (D1, D2, D3),
all in the ages between 13 and 15 years and after their period of
voice breaking (audiograms of the deaf boys, see Fig. I-C-1).
The sentence used in this analysis can be transcribed
in broad transcription as /suglen finer men de:kalt/.
Fig. I-C2 shows sonagrarns with segmentation of this
sentence spoken by the three normal-hearing boys (H11 H2, H3).
In these sonagrams one can see that the five words are spoken
continuously without pause, except for H32 where there is a pause
between " skiner" and "men". "Det ar" is spoken in a colloquial
manner as /dera/ or only /deg/. Speaker H2 has a rather reduced
articulation, he speaks a Stockholm dialect used by teen-agers.
113 speaks very well; well articulated and clear, but in a col-
loquial style. A segmentation of the utterances of these normal
subjects is carried out in Fig. I-C-3. In this symbolic repre-
sentation the time lengths of the segments ,are not shown. A mean
pattern for those three subjects here referred to as a normal
pattern has boon constructed and is shown in Fig. I-C-4. A dis-
tinction is made be set anent and features of primary impor-
tance and those of secondary importance. The individual pattern
of speaker H3 comes closest to the normal pattern. Fig. I-C-4
shows the segment typo feature pattern for the normal version.
Description of the ;e ;menu of the normal nl` -1-ern
Ser,vent_l pertains to /0/. Its sola'ce feature is noise and the
resonator feature is fricative.
-7'
12.3
230
21.3
1
D 1930
WO
OW
OO
LCC
O
I1
1ii
I
'
1!ta
;
1,1;
t, ,,
;I.
10
.!!t
II
':1
i10
1
.77-
7.77
.77.
"71
20t
3,
;i
1 r'
11-1
TH
',H
II--
r-t
1--1
-1-1
-1-1
--;
77-7
13,
1-7,
---.
.- r
---
r-t.t
---r
----
:--
-1-2
--i-
s.ri
- -7
-r-T
T!
ii
;
4.4.
,
ii
,-E
ri
1.
.
1
ii
'
1
1-7-
1 -T
-1-7
:S
3.."
-7-T
'''4-.
"1--
-\--
411_
,c. I .!
300
HO
30 30 40 70 SO 93 19
3
;-k
,,,,.
!4,
..i
t_i
t1
.!
1___
tI
i,,_
_I,
1--t
---7
-1--
i---
ts::.
,1-
i -.,.
,,i
1
1---
-niT
-1,-
1-7
--1
,,
,,.
-,:,.
.14
..
1:0
--;-
-.--
;1-
-1-]
1/0
-17:
1111
I
ir. L
1*,
t.,11
,.:..
1-1.
11_
,_,
.,,,
!i
?!
!!
,L.s
,I_
LL
Ii
1L
ig1
1,
-,,,,
,,,_,
-:
it___
!,
102
Ili
1111
Ilif
.i. 1
,:
1'
.I
130
i 130
FRE
QU
EN
CY
IN
CPS
-o 77
- 30 - l0 00 20 30 40 50 00 70 S
O 90 100
110
12o
130
123
2:0
300
u214
4440
:000
4000
$000
1, ,
...,
L_
,I
1111
1111
1111
11
ii
i1
j I.-
--1
!
i;
,,H
-,-
--,
i,
n,.
,
I_.
......
._.
i,,--
!--;
-1i
i1
i-r-
1i,
iiI
lit 1
1;si
iiii
ii
Illi!
-ri
III
1,j.
11
:71
-1l
-1-0
I
'
'Lif
t!!
.L...
..L...
:il
i1.
imam
1
it.,,.
it:
1
Iit°
1I .
i11
.i
1 .I
i-1
iE
n I
t
7-1.
-1.
'.I
Itj
I7-
71-T
iii
Li i
:1.
1
,i
t1
..._.
,_.,,
..
11H
-11.
.L
-'''-
--...
t--r
-1,
11
it
I'
,''s
c.1
11
I
, 1:1
:It
!:4 EI
----
t_I
1
v .
t1,
II
NM
11
7.2
N.
,f
1L
ii
i1
80,
1.
:N.!
iI.
ri"
a.rr
'i_Q
:I
!""
II
1
I.
I--
-',7:
..,11
4.4,
'.1
I,_
._..1
..9
o
___I
_11
2`IC
4,, I
...21
.;
,
1L,-r
-i-!
,.. t
--..
--i_
__.,-
-2--
Li
.te
NM
I,
:
,,x,;;
;1--
.-1-
-,,
--;-
10-,
,-;
,1
1!_
;i .
h_...
.)i
Iez
'-'1
",:
1us
I,_
_L
_Lf:
-::.
1j j
° I_
j 1-
%--
-H°.
_1
t
III
rnu
i'
II
I20
1111
,!-0
-1i
11
I1
INM
IMI
MIN
.
FRE
QU
EN
CY
IN
CPS
20 111
10 20 30 441
30 I0 71)
SO 40 100
110
120
120
Fig.
1-C
-1. A
udio
gram
s of
D1,
D2,
and
D3.
Air
con
duct
ion.
Rig
ht e
ar =
sol
id li
ne, l
eft e
ar =
dot
ted
line.
-20
-ID 10 3. so 40 30 70 $0 100
110
120
123
230
300
D3
1000
200
4003
1000
-'...
..
1I
I1
1t
__i
1-1-
i---
ti--4
,-_i
_-_,
..1-
,1
1
-.7-
t-1-
1--
ti- !
-'0-1
--1-
'
.,-
-iE
l Iu,
--r-
--1-
_..i
1___
I___
i_l_
t.-1-
',-.
I
- e
-- :-
-t -
. -1-
-I
1
1
''.
1i
1.-i
i:
4
., .11
-1t-
-;i
--i-
--',-
-to
--,-
-1 -
4!
.
I1
;---
- .4
-41
- -i
IMII
---t
iI
1,_
_.7-
1 :-
-.. -
4,1-
i20
1
z--
--4
-1 -
----
; - -
r-
i;
'4
.4
!I ,
!
____
__1_
-t-
.
____
__i_
__I-
.1
I'
i'
1i
1:
I:
EN
--14
4-1j
:--
--itt
-'I
I,
40
- i -
I - -
1,-
- i
,-
1.-1
-:`
--i-
1.
.1
510
130
:1
--4
--4-
- i
1t
it
1
1--
i--.1
- i
; '';
H.4
.'-
I-
4'
1!
10i
I1
f
___
.t
-I-
/;
:t
1
i__
70 ..
... -
-
,i
1.
1--
1.:N
.1.
,I1
1_1
_4_1
_ .
.....f
__...
i.4_
.1
sot
,. --
rya
,
N'.:
4-I-
-'
-4-L
it
I,
I,
II-4
!
i---
-14'
,. i -
-1 -
;---
4-
I.--
-- -
t-
i1
!
sisc
?o.r
."5.
..r-i
--:-
t-,-
..-,..
..-7-
1-1o
ei
#,
":
:
-4 -
-L :
I
-"I-
i 110
-4'-
.--.
_ t
-.II II
IMI
I,
1I
II
'I
'lit
44 -
-:i
'41
r-1
lI1
1
,
120
412
0:
##
1M
ON
--I
--
t II
III-
1iIN
I
FRE
QU
EN
CY
IN
CPS
kc
7
6
4321
4.
3
;, ') /
4 itI
iflit
2 :5 .1,
L/1:3 3 1 i Ili 13 t1.,../111117 2 0
6 7 .8 6 10 11 12 13 1.4 15 i t3
!<,,;
7-
i
4-.
2__.
2---
1_.
0--I
I I
1 11.,;11 1141 ;
--r-i I OM1.". -7. J-1-1 1.11111
I- , I, ;: 3 Q311 3219 11/.1! 2.)1
1 j J.__ _I3/1---T-r-r-r-r-r I, f f
0 .1 .3 4 7 .3 10 11 12 1:3 o3 1.9 10 21 2.2 sec.
Fig. I C-2. Sonagrams of the test sentence spoken by the threenormal hearing subjects, H1, II2, and 113.
r.
HI
Sezi
men
tsz
Voi
ceN
oise
Tra
nsie
ntO
cclu
sive
.Fr
icat
ive
Lat
eral
Nas
alV
owel
like
Tra
nsiti
onal
H2
S;:o
rnen
ts:
Voi
ceN
oise
Tra
nsie
ntO
cclu
sive
Fric
ativ
eL
ater
alNasal
Vow
eiti
keT
rans
ition
al
H3
Segm
ents
:
Voi
ceN
oise
Tra
nsie
ntO
cclu
sive
Fric
ativ
eL
ater
alN
asal
Vow
eltik
eT
rans
ition
al
11 i
10fl
__
_
i
_ _
.._.
!'
1
,,
x!x!xix xi
xx xixlx!x x xix x,
;I
x x
xIx
ii
!ix;-
1
'A- x
A..
X
11
XU A
X;
XI
Xi
-1-
X X
IXX
;ki
VA
Ai
!X
iI
X.
lc, A
Ii
!IX
XX
XX
XIX
<T
rx
xlxy
,x;
xx
ix!x
-I
xi
xix ix
k
X x aI.
x ix t,
xlxi
xI
xx :
,x;x
i1
n I
E.
irni
8n
ixi
d -r
e:sl
uili
E I
nf
12
3 4
5 S
7 8
910,
11 1
2 13
,14
15,1
6 17
,18
19 2
0 21
22
23 2
41
!:
,!
1;
;!
I!
I!
!:
,,
1
X;X
XX
iXX
'XIX
IXI
iX
:
X
IiX
;
xjX
i
I: X
Ix'
i XI
X1
XX
Ii
I'I
1 x
.x;x
;x x
;x,x
x!x
X X
!!X
X!)
, CX
:X;
XiX
:X
,X
I
[ k
IX X x Ia
li
tx;
I>L
__!x
Ixix
;'A
Ixi
SI
u: 1
,-1.
I E
Ih
ii :
iri
ffE
1 m
l E I
n 1
e:I
8 9
0313
2333
41
III
Ili
111
;I i
1
II
iI
ilX
X, I
.--
--4
f1
X : X X,X ){___L
,XIX X)Y!X1X1
XXXXXiX
1>71:-X
ii
x 7
xr-
--;
Ix1
i
xiX
XI
,j_
___t
XX
XI
Ix;
I N
.I
1/ I
'/%
!I
xx
xX
1x
i,x
4---
--1
X;
x !
,x
XI
.i
I/
1
11
!
X 1
xX
X1
X
tA
II
[X X
;XI I
X X
iX
X i
XX
XX
! X
X I
X
i---
--4.
.....1
1X
Xx
xx
XX
XI!
X
1x
II
XX
IX
X111
XX
11
SIt
.i:I
II
E.1
is1
n 1
81r
ofri
en
de:
ka
I
Fig.
I-C
-3. S
egm
ent f
eatu
re p
atte
rn o
f th
ete
st s
ente
nce
for
the
thre
e no
rmal
hea
ring
sub
ject
s 11
1, H
2, a
nd H
3:
Seom
ents
:
Vo
ice
Noi
ser
nSie
ntc
LU
IS v
e.
Fric
ativ
e'C
. '2
Fa L
NC
:SC
A
CA
VZ
1_1
e
iran
sitio
nal
1
I
2 3
I , ,
45
57.
8 9
1011
1213
14.15
1617
1819
2021
22f23
2425
2627
2829
30
./1
Av A
V AV A
v AV A
V "VIN
00
I A0
XX
X\ i
A,Af
0X
Nm
e
0N
.,0
XV A
X
N./ A
V!
.9 A!, A
0V A
v A0
v AA
XV
.X
XX
VV A
v AV A
, AAi
V AX
00
XX
X
Af
AV A
v AV A
V AX1 _
0A
N,/
0X
Sq
Li:
1 If
1:n
E..,
,...,
....
.
ka
Lt
Fig.
I-C
-4. N
orm
aliz
ed s
egm
ent f
eatu
re p
atte
rn o
f th
e te
st s
ente
nce.
18.
S2samic? pertains mainly to /s/ but it is influenced by the
fel' 'ng /u:/ Its source feature is noise and the resonator
features are vowel-like and frictivo. The vowel-like feature
is common at the boundaries between fricatives and closed vowels
and is realized by the appearance of F2. The friction is reduced
and at the last 10-15 inset it totally disappears.
Sep*ment 3 pertains mainly to /ut/. It has the features voice,
vowel-like, and transitional. The transitions at the beginning
of the /u:/ are influenced by the /s/ and at the end by the fol-
lowing /1/.
Segment 4 pertains to /1/ and has the features voice and lateral.
Segments 5 and 6 pertain mainly to /c/. They are voiced and vowel
like. Segment 5 is transitional and 6 is both transitional and
nasal. One might prefer to include a stationary part in the
middle of the vowel, but this did not seem motivated here. The
nasalization can start earlier. In that case segment 5 falls
away (see H2).
r3ornent 7 pertains to /n/. It has the features voice, occlusive,
and nasal. This scEsment is not important for the perception of
/n/. A strong nasalization of the preceding vowel is a sufficient
cue. This segment is missing with H3 and is very short for H1
and H2.
r
Sof,lonts 3 and 9 pertain to /1/. The /0 starts directly with
a segment (8) with the features noise, fricative, and finishes
with a seEAent (9) with the features noise, voel-like and, frictive.
In case the allophone[q] is used as with subject H2 the feature
vowel-like coexists in both segments and at the end there enters
the additional feature transitional. For speakers H1 and H3 1
there is an interval without friction before the vowel of about
20 meoc length. Usually there exists such a segment in the f
boundary between 07 and /iqbut much shorter and of no impor-
tance.
Sef.ments 10 2.nd 11 pertain to /i:/, The features are voice and
vowel-like. As eegment 10 is influenced even by the preceding
/17 there is the important additional feature transitional.
Socjment 11 is inCluenced by the following /n/ and has therefore
the additional feature:; nasal and transitional..
19.
Segment 12 pertains to /n/ and has the features voice, occlusive,
and nasal.
Serments 13 and 14 pertain to /c/ in "skiner". The first segment
(13) has the features voice, vowellike, nasal, and transitional
because of the preceding /n/. Segment 14 is not nasalized but it .
is transitional. For speakers H1 and H3 the phoneme /r/ in
01:neri is highly reduced. It is indicated mainly as a special
formant transition in the preceding vowel. In case segment
15 (N) and segment 16 (pause) are missing, in accordance with
their secondary importance, the nasalization goes through the
whole of the vowel /E/ which is so reduced to one segment, see
speaker Hi (Fig. IC-3). Therefore segment 14 of the normal
pattern is of secondary importance only.
sermalul pertains to /r/ which in this combination is a voiced
fricative variant. It has the features voice, noise, fricative,
and vowellike. This segment is not of primary importance.
Segment 16 with the only feature occlusive stays for pauses.
It is not an important segment.
Se,csylent 17 pertains to /m/ in "mon" and has the features voice,
occlusive, and nasal. For speaker H3 this segment is of the
order of 10 msec length. This is unnaturally short and only by
the strong nasalization can a normal perceptual effect be achieved.
Se.gent 18 pertains to /c/. It has the features voice, vowel
like, nasal, and transitional. In the second part of the segment
the formant transitions are quite slow. The segment is nearly
stationary and it could perhaps be subdivided into two parts,
the second one lacking the transitional feature.
Sei7ment 19 pertains to /n/. It has the features voice, occlusivo,
and nasal.
Segments 20 and 21 pertain to /d/. Segment 20 is the occlusive
part with the segment features voice and occlusive. This segment
is related to the interval of velar closure'before the opening
of the oral passage. In spite of its short length this segment
is important for a correct /d/, since it allows for the over
prOssrre preceding the explosion of the stop sound. However,
it can be difficult to identify and is attributed a secondary
importance here. Segment 20 rs found with the speakers El and
ti
20.
H3 but not with H2. In case of H2 oven the following segment,
i.e. the explosion is missing. Segment 21 is Ihe explosion of
the stop soetnd with the featuree voice and transient. This
segment is quite weak or missing, if the occlusive part (seg
ment 20) is too short or missing, as was the case with H2.
Segment 22 pertains to /ege3/ or /eg/ depending on the verifi
cation of "det ar ". The segment has the features voice, vowel
like, and transitional. The transition at the beginning of the
segment depends on the preceding /d/ and on the end of the
following /k/.
Segments_21 and 25 pertain to /k/. Segment 23 with only one
feature (occlusive) is the occlusive part. Segment 24 is the
explosion with the feature transient. Segment 25 is the aspirative
part with the segment features noise, fricative, vowellike, and
transitional. The /k/ is here in word. initial although phonet
ically intervocalic position. Intervocalic, voiceless explosives
have usually a short aspirative (vowellike, n)ise) part lacking
the features frictivo and transitional. The /k/ is in the present
case in word initial position and the aspirative segment is longer.
The features frictivo and. transitional exist but appear to be of
secondary importance.
Segment 26 pertains to /a/ and has the features voice, vowellike,
and transitional. In the middle part of the segment the formant
pattern is nearly stationary.
Sepment 27 pertains to /1/ and has the features voice and late2al.
ler;monts .28_,291_and30 pertain to /t/. Segment 28 is the occlu
sive part with the feature occlusive. Segment 29 is the explosion
with the feature transient. Segment 30 is the aspirative segment
with the features noise, fricative, and vowellike. These are
the common segment features in a postvocalic, final /t/. Because
the iti here follows a lateral the aspirative segment has a very
low in and is barely visible. It has therefore been con
eiaered as of secondary importance.
Comparisons of 4ho def subjects' speech with the normal pattern
Sonagrams with segmentation of the sentence spoken
by the throe deaf subjects (D12 D22 D3) are shown in Fig. I - -C-5.
Fig. I-C-6 shows the segment type features of each segment.
AC
;i;
1
i-1:7
i?.1
1,4
15
5 :7
;8:S
:1G
,:12,
131!
151j
3; 1
7.;
'2 5
12
"1/4
-,2'
'2d
2S,
II
11
;,
.
.i
I,
..
3,
3;
I*1
t
12
1.0
IL;
l.2;3
4-I
L1:
.31.
71C
.s.
1.9
20 2
2.2
232:
42
3iT
;1
1I
...
.,
;;7
,-4.
. r.:
i..!
;;
;:
;I;
.1.
t.
..
,I
;!
. L.
1.
3.;
.4a
.1;.
3 1--
II
;
4 15
1 6
17
18
i.110
;11
1141
;,15
:171
1120
;;
22 :1
24H
26'
'281
2911
1341
39i 3
637
381
39;
Li;
11t
,44
;11,
,ⅈ
4.4i
41)1
41,i
ftL
12.
3.4
.5.5
.7:3
1.2
1.3
.171
2.0
2.1
2.2
2-3
2.4
252.
627
2.8
29ao
31
32
3233
1I
361
1
,i.i
I';e!
.11
,
2.5
2.7
2.8
2.9
3.0
3.1
3.2
3.3
3.4
3.5
3.6
33 3
9 3
4-,
1.1
;
38
D2
...1.
1111
1111
111
Fig.
I-C
-5. S
onag
ram
of t
he te
st s
e-ite
nce
spok
en b
yth
e th
ree
deaf
sub
ject
s D
I., D
2,an
d D
3.
sec.
'
D3
,;(
Di
D Seg
men
ts.
1,2
3,4
5 6
7 8
9 10
11
12 1
3.14
15
16,1
7,18
,191
20,2
1122
23.
24,2
5 26
,27
28,2
9 30
31
32 3
3 34
35,
36,3
7,38
,39,
111
;;
1i
1;I
i,
1.
i'
I;
1I
;;
I
;i
I
;;
I
II
i 1i
x x!
xix'
x:xx
i:x
xjx;
x )(
Ix1
,xi
xiX
Ixi
xx
x;I
1i
i.
,i
1,
1i
1
X,
iI
X X
;'
11
i!
II
X!
Voi
ceN
oise
Tra
nsie
ntO
cclu
sive
Fric
ativ
eL
ater
alN
asal
Vow
ellik
eT
rans
ition
al
D2
Seg
men
ts:
Voi
ceN
oise
Tra
nsie
ntO
cclu
sive
Fric
ativ
eL
ater
alN
asal
Vow
eilik
eT
rans
ition
al! 4-"
D3
Seg
men
ts;
,
.
x;
1
i
,XI
i
i x-T
TI
ii
i i x!
Iix
I
xx
xiix
x;
xI X
.
xi x
X;
;I
ix
'
xi
1
i.
i
x : x
I i
;x
xi:6
1
1
; I
I ,i
.
.x
i.
i!
i--1-
-I
[1
i .x
i I
I,
X ;
X X
I X
X j
X X
-L-
ni
' I;
I X
X i
XX
X!
XX
1 X
! X
IX
I X
' X!
;'
X X
!!
X;1
X ' ;
XL
.--:
-;
----
1.--
-X
! X
j X
,X
'1
i'X
j X
1 X
-1--
--X
;lx
lXX
X
X ,
X:
X '
I X
!!
X ; i
X !
X[
XI
X!
! X
1,
i
,X
i I44
f k
al, 1
tir
--t-
C--
-s
I11
:.
1C
i.
-n
.-4
5i
I:1
n6
r 1-
41m
, El n
114
d I
e:
I,I
1";
!:
',
I1
1;
,'
i--
----
t,X
X: X
X .
IX
I X
!.
'' X
I X
X ;
II
X X
X I
I
X'X
50-
--I i_
-_,
ix X
!X) i
--!
1
'1
X.
: X !X
;;
! X.!_
_.I
Ix;
X, X
.,
1,
' Xi
I!
IX
IX
I,
----
..--,
;i
!X
j X
,IX
'!X
; Xv A
1
1 1
X!
X
..
.__!
r--r
.i
:---
-1X
:X;X
Ifr
1,
i x i
Xi i
;X!
_i,
L__
_iI
!1
!1
;I
X;
X X
:X1
!:
iI
i
..;
,,
____
ti
!f
-tI
i-.-
---!
!
!XIX
'X
:i
1,X
XI
X;
11
'X;X
ii
X'
i'X
XI
!.
!,X
j XI
XI
"XI
XI
XI
XiX
:X1,
.!X
,,
I
''
I!-
--.-
'----
-,
i..1-
17.1
..'.
Voi
ceN
oise
Tra
nsie
ntO
cclu
sive
Fric
ativ
ee:
oral
Nac
cl.
Vow
elM
eT
rans
iticn
at
L
1:1
ni
rE
ne.
i.
"e
a I
tI
1 32
33
34.3
5 36
37 3
8 39
.40
---
- I
-- -
- -
-,
I'
i II
!'
1
1
1!
1I
1I
I1
-Hi
.i
-r-
-1-
'---1
i4
X X
tX--
11i
X;X
X X
IX:X
X,
)(1)
( X
X X
X X
Xi
X X
X'
___I
XIX
X,X
X X
XX
;X x
.t
X!X
iX1
X!
Xi
XX
.I
XI
r---
--
!;
r.
i
';---
i:'1
L1
X1
iX
; X
XX
XX
X1-
7
X I
XX
XIX
IX
IX
X
xr--
-+I
xI
1I
i
Ix
x(x)
x x(
x),
IX
x x
x:
x x
x:4
Xx
x x!
'X
X X
wX
X X
XI
XiX
XI
xX
XX
X X
XIX
X x
'X
Xx
x
s.
i
L:
i.6:
n I
1:in
'c.
:.r
0 M
E n
IT.4
de:
6...a
.*k
ai
it
Fig.
I-C
-6. S
egm
ent f
eatu
re p
atte
rnof
the
test
sen
tenc
efo
r th
e th
ree
deaf
sub
ject
s D
T.,
D2,
and
D3.
21.
Speaker D1
Fig. I-C-7 contains idle sonagram and diagram of
evaluated segment type features of subject Dl. The correct
features present are marked with x. Features in the normal
pattern denoted as primary and of secondary importance are
included as correct. The present but not necessary features are
marked with 0, The present but erroneous features are marked
with yal . Missing features of secondary importance are marked
with o. Missing features of primary importance are marked with
Segment 4 of D1 pertaining to [us] departs from the
normal pattern by being stationary and nasalized. The misplaced
nasalization extends throughout segments 3-7 and the major part
of the sentence. The nasalization is a real problem for this
speaker. The junction of [n] and [I] (segments 10 and 11) shows
:Irregularities. Segment 10 is a misplaced interval of dovoicing
due to abnormal rhythm but is not a serious error. [J] starts
with a transient segment which is erroneous. The voice onset in
[is] is delayed, resulting in an additional segment (14) with the
features, noise, vowel-like, transitional, but also fricative.
Without frictionithis segment could be the starting transitional
part of the corresponding segment (11) in the normal pattern.
However, because of the friction, it is related to the [f]. In
segment 15 the transitional feature, which is considered necessary,
is missing. The junction [n]-[c] in "skiner" is erroneous too.
There is an occlusive segment (18) and an explosive (19). The
control of the vocal cords and of the velum is not correct here.
This results in some kind of [d] between [n] and [6]. In segments
22, 31, and 32 there is an erroneous nasalization. Segment 21 is
an unimportant stationary vowel segment. The /d/ in segments
29-30 is voiceless and [1] in "kallt" has both voice and noise
as source features. The aspirative part of [t] is missing but
this failure is not of main importance.
There is no missing segment feature transitional
(except in segment 15), but the transitional segments are quite
short compared with the stationary vowel segments. So, for
example, in segment 36 we have the feature transitional, but the
transitions in some parts of the vowel are quite erratic and
show that the articulation is not really under control.
Ic
1'7
;iz
i; 4
;6
1714
41J
2027
2329
:
32ln
3638
1i
;j
j
.t
.i
Jil
lit5
.(7)
it9
1.3
1'.`
3it;
17-"
31'
92:
321
2-2
2.3
2.4
2.5
2.6
2.7
2.8'
3'.0
31.1
333.
43.
53.
63.
7 38
3:9
40
4.1
sec.
Seg
men
ts n
r.D1
Nor
m.
Voi
ceN
oise
Tra
nsie
ntO
cclu
sive
Fric
ativ
eL
ater
al.
Nas
al.
Vow
eilik
eT
rans
ition
al
12
3 4
5 6
78
9 10
11
12 1
3 14
15
16 1
7 18
19
20 2
1 22
23
24 2
5 26
27
2829
30
31 3
2 33
34
35 3
6 37
38
39
71-
4I I
1
Ale
I,iii
1
\of
II4
.
11
11r
1
war
,/,".
11,L.
,IS
=dm
,.0,
Om
%I
e.,o
r O
mrm
.,/
/
Om
,/
IxE
xix
xix
xii
ix
xx
ixai
xxx
xoex
mx
xx
x1
II
xix
',.1
E1
.?',,
xx
xx
io1
;
ixi
El
ix/
/,..
xx
xx
xx
xlI i
i
xixE
,Li
xI
Ix1
7/ 7 7
xx
i 1D
I 1
xx
x>
:y
ix
xx
x 1
x 41
-''x
Li
xix
[5_,
:ixx
x2,
xx
xx
xx
0xi
x1x
x1
mo
xx
xx
xo
xS
iu:
l2
n.f
Ii:
nE
rilr
nEno
de:
#kal
t
Lege
nd Pre
sent
inco
rrec
ted
segm
ent f
eatu
rLID
Pre
sent
, not
nec
essa
ry fo
r id
entif
icat
ion
XP
rese
nt0
Abs
ent
0 A
bsen
t, ne
cess
ary
Fig.
I-C
-7. S
onag
ram
and
eva
luat
edse
gmen
t fea
ture
patte
rn o
f th
e te
st s
ente
nce
spok
en b
y D
I.
0,, 0
},
Pr.
.
The nasalization is, as mentioned above, a problem
for the speaker. Voice source quality and pitch range are quite
good.
Speaker D2
Fig. I--C-8 shows the sonagram and the diagram of
evaluated segment features. The initial fricative is of infe
rior quality. It is aspirated the whole time and not only at
the end. (The fricative spectrum is not correct either.) In
[1] the vocal cord vibration stops. There is a silent segment
(4) whicth is followed by a segment with noise as source feature.
Both the segments (4 and 5) are incorrect. In the following
vowel [ e] in segment 6 the important transitional feature is
missing. The whole vowel is nasalized. According to the normal
pattern it should not be so, but this is of secondary importance.
A lack of transition can be observed in segment 10, i.e. in [i:]
too. The junction [n][E] causes some difficult ins even for this
speaker as is the case with speaker D1. There can be :en a
voiceless occlusive segment related to the velum closing and the
vocal cords set to rest before the oral cavity opens. The release
is a kind of explosion with a following aspiration. A [ti can be
heard because of these irregularities. In the [c] in "skiner"
the features transitional and nasal are missing which both are
of main importance. The transitional feature in the [6] in "men"
is an important, missing feature. The /d/ in "de" (segments-22,
23,24) is voiceless, not only the explosion is voiceless but
there is also a short aspirative segment. The result is that
/d/ sounds like a [t]. In [e] and [al of "de" and "kallt" there
aro stationary segments of no importance.
The main difficulty for su'oject D2 is the improper
control of the vocal cord. functions (see segments 4-5, 12-13-14,
22-23-24). The vocal cord vibration 'seems to be too weak. This
is the reason for the numerous fall out. A second difficulty
exists in the failing transitional features. The speaker has
a nearly normal speech tempo and rhythm, but the right speed does
not always result in a correct transition from one phoneme to
another.
O
segm
ents
nr.
D2
Nor
m.
1
Voi
ceN
oise
Tra
nsie
ntO
cclu
sive
Fric
ativ
eL
ater
alN
asal
Vow
ellik
eT
rans
ition
al
0C
;3
'7i3
O"
;I
"7-1
;.
tI
311-
t 1,iis
.iiii
1819
2.0
212.
2se
c.
32i
1341
351
1.5
1.4
1:5
1.6
17
12
3 4
56
7 8
910
11
12 1
3 14
15 1
6 17
18
19 2
0 21
222
3 24
25
2627
28
2930
313
2 33
34
3536
2124
257S
972R
2c11
01
.L
I1 Ix
xx
xxxo
eIx
-0al
____
__1
____
____
_
!xi
xi1
!x
xi1
axj
x!
ax)(
le,
Ox!
'1,/)
,I
)(ix
7,X
i{
1i
xx
xx
xX
ii
1
0 x
1I
Ix
x.
I-''
-7),
I
II
Ix
xx
xx
,Xx
Io
xx
oX
Xx
EN
xo
xx
x00
000
0x
lao
Y.
su:
1E
nj
.n
f-rm
En
e:k
al
Leg
end
Pre
sent
inco
rrec
ted
segm
ent f
eatu
reP
rese
nt, n
ot n
eces
sary
for
iden
tific
atio
nX
Pre
sent
0 A
bsen
t0
Abs
ent,
nece
ssar
y
Fig.
I-C
-8. S
onag
ram
and
eva
luat
ed s
egm
ent f
eatu
repa
ttern
of
the
test
sen
tenc
e sp
oken
by
D2.
Speaker D3
Fig. I-C-9 shows the sonagram and the diagram of
evaluated segment features. The diagram shows that this speaker
has the greatest number of errors. The voicing after a very
weak [s] starts only slowly. Consequently there are both voice
and noise as source features at the beginning of tua in segment
3. The noise feature is incoriect. The other segments of Lu:1
are transitional and stationary alternatively with erroneous
transitions. The nasalization starts too early already in the
middle oP [u:] in segment 5. Probably the [1] (segment 7) is
nasalized too.
23.
The nasalization continues in the following vowel
[c] which is not incorrect. The important transitional fea-
ture is missing there. D3 has difficulties with the correct
timing of the onset of the vocal curd vibration at the junction
between [ i] and [i:] as well as at the junction between [s]
and Ng] (segment 3) and accordingly segment 12 appears as a
voiced fricative. The whole segment is incorrect. In [i:]
there are some minor errors owing to missing or too short tran-
sitions. The formant transitions are in the following vowels
missing in the two segments 17 and 22.
The burst segment of /d/ of "dot" is a short voice-
less fricative instead of a voice explosive. D3 is the only
speaker in both the hearing control group as well as in the deaf
group who says "ar" A'ar/ instead of the normal reduced form [e]'
or [e] . There is a nasalization in the segments 28 to 33. The
[r] is rolled with very fast occlusions (maybe uvulas). After
rr] there follows four unwanted. segments. The whole utterance
"dot ar" is realized as rds;)rGno], spoken very fast at the end.
/k/ contains a back fricative [x] and the explosion is missing.
In segment 38, i.e. [11, the source is incorrect, there are both
voice and, noise as source features.
The pitch of speaker D3 is very high and he has
difficulties in joining voiced with voiceless sounds. The ar-
ticulation is often incorrect too. There are long stationary
vowel segments with only quite short transitional segments.
All the stops of D3 are incorrect.
kc7
Seg
men
ts n
r.D
3
Nor
m..
Voi
ceN
oise
Tra
nsie
ntO
cclu
sive
Fric
ativ
eL
ater
alN
asal
Vow
ellik
eT
rans
ition
al
2 1 ;.1
b
,;
,)
f iT
61
2.4
"
I;
,-1
---1
-'1'',
3,'
i;,--
,!-.
11!-
.:iit'
kr7.
"..r
"..-
'.1!
..i.-
,r.
ii 9/
1,("
.'t-
.)°
29 '
;311
'13
(135
1 36
1 3/
sbi
39li
::;..-
.;:_:
_,,
.....;
,,,,
;Gv.
:.. :,
...L
....
...:,`
......
.{L
...,.
1iii
;1
e1
i
i;
lii
rli''
'',,
",'i
j_L
L:
i;!
.___
__,H
i!I
.1i
iiI
!,1
:, ti
tf
__I:
ifL
iliie
llilli
i:sii'
li.6
-:.(
-)u
1.2
1,".
-it.,
..:-;
1617
1:3
10 2
0 4.
22 2
3 24
25
2.6
23 2
.8 2
.93.
0 31
32 s
ec.
12
3 4
5 6
78
9 10
11
12 1
3 14
15
16 1
7 18
19
20 2
1
12
34
56
78
910
11 1
2 13
14
15 1
6 17
22 2
3 24
25 2
6 27
28
29 3
0 31
32
33 3
435
363
7 38
39
40
18 1
9 20
21
.,
1
11
1I
1
,IX
ti ii
7 x
10: x
x,
z.,
(2/1
x !
Ffi x
xx
x x
x x
x 0
el,;-
--i-
----
-,4
___
7-7
,,-;',
,,
;/ii
1!
x x
x2
1
;!
;
!;
:'
!
ii
1 x
xi x
rx
:i
i
i!X
X,
>:
,1,
;
1---
;:-
-;%
.IX
.X"-
--1
i:tr
YC
-IX
X%
-,, /
!;!
t/ A
''.
.ri
IX;O
ICIT
XX
11X
1'
iOiX
/'
211
ii
! X
i1
iPi
®!
1 el
is
IU
`I
L i
E I
ni 1
i:
Leg
end
Pre
sent
inco
rrec
ted
segm
ent f
eatu
reP
rese
nt, n
ot n
eces
sary
for
iden
tific
atio
nX
Pre
sent
0 A
bsen
t0
Abs
ent,
nece
ssar
y
23 2
4 25
26 2
7 28
29 3
0I_
11
la X
xI
Fig.
I-C
-9.
Sol-
v(1g
ram
and
eva
luat
ed s
egm
ent f
eatu
repa
ttern
of
the
test
sen
tenc
e sp
oken
by
D3.
24.
Summary and discussion
The compaxison made between the speech of deaf and
normal-hearing chilaren reveals he main difficulty in phonation
and specially the la.ck of synchrony between articulation and
phonation, e.g. at the junction between a fricative or a nasal
consonant and a following vowel. Prominent errors in velar
functions have been observed in the form of abnormal and mis-
placed nasalization of vowels but also a too early denasalization
of nasal consonants.
A third chal.acteristic apparent from the study is the
superstationary form of vowels. Transitions are often missing
or if present they are too short indicating too abrupt articulatory
movements. Errors in speech rhythm were not evaluated in this
study since additional measures would be needed.
A critical evaluation of the segmentation techniques
employed in the present work reveals certain ambiguities in the
labeling of segments and the determination of their boundaries.
However, v9,riations in the detailed segmentational transcription
of the material would not have effected the conclusions of the
study.
A further development of the techniques for articu-
latory interpretation of spocrograms and perceptual evaluation
of pattern abnormalities by mean6 of synthesis experiments is
planned. A study of feature distribution in non-normalized
time is planned For a coming report.
Reference:
(1) Fant, G. and Lindblom, 133: "Studies of minimal speechsound units", STL-GIPSR 2/1961, pp. 1-11.
B. STUDIES ON THE SPEECH OF THE DEAF. II
J. Martony.
31
In the first part of this study (published in STLQPSR
3/1965) a speotrogram segmentation technique has been described, On
the basis of the segment feature pattern we have an objective tool
for the description of ap utterance and of its possible abnormities.
Anyhow, the method has some lacks. By definition the -
segment boundaries coincide with a change in the feature composition.
An exact determination of segment boundaries is very difficult in
some instances and by slight shifts of the timing of features new
segments can arise which are of short duration and usually unimpor
tant. The other disadvantage of our earlier approach was that the
time duration of the segments was not indicated.
A representation of features as a function of time is
perhaps more :advantageous. In Figs. IIIB-1 to IIIB-4 such feature
patterns are shown. The utterance and the speakers are the same as
in the first part of the study.
Fig. IIIB-1 shows sonagram and feature pattern of the
test sentence spoken by the hearing subject H3. Test utterances can
almost be considered as a normalized feature pattern except for the
pause between the two parts of the test sentence. Figs. IIIB-2 to
IIIB-4 show sonagrams and feature patterns of the test sentence
spoken by the deaf subjects D1, D2, and D3. These sentences were
discussed in the previous part.
A comparison of the length of the segments or the features
for the different speakers is not easy because of deviations in the
number of segments and features. An alternative method is to study
phoneme durations as shown in Fig. IIIB-5. Of course a determina
tion of phoneme boundaries is not unambiguous and to avoid large er
rors in some instances adjacent phonemes have been taken together
in groups. Generally as in Fig. IIIB-5 it is apparent how much
lower the rate of speaking of the deaf subjects is oompared with
that of the hearing subjects.
6 5
H3
su:L
Enf
i:n E
r4t
mE
nde
k al
3 2 1 0S
egm
ents
no
----
- V
oice
Noi
seT
rans
ient
Occ
lusi
veF
ricat
ive
Late
ral
Nas
al.
Vow
ellik
eT
rans
ition
al
-r4
tt
VT
!, 4
1..7
.004
111e
Pig.
lnyr
1.,
ii:it
n 1
I rj
lifo:
Z:,i
i,:ia
i;Ida
it?`
..-
ik.,
7,k
tthi,,
;1i,1
,j'-
iii:p
orl-
k.
_...
...,.
_Ii
,U11
1:it!
iiIII
P'
,LA
.,14
i1;?
.;iiii
;t.i ;
ill!.
..
I R
I
.,1I
igiiq
iimi!
i:kiii
::::..
r_ ..
.,:
...--
-.,
,...-
_,.,.
.
13
410
1114
_
1921
2530
3132
34
,___
____
,.....
......
...._
_6.
,M
IMI
It
MIIM
'11
4---
-141
16,-
--.1
1111
0
MI
ND
ON
IIIM
IIMIII
IMel
,
.41=
04.-
......
.-..4
1111
1111
1111
10--
--41
1M11
MIM
EO
l11
1011
111
Ici
.12
.3.4
III
III
.910
1112
I1
IIL1
i'i'l
i,13
14l 16
11is
192.
02.
Fig.
III
-B
-1.
Son
agra
m a
nd f
eatu
re p
atte
rn o
f th
ete
st s
ente
nce
"sol
enak
iner
men
det
fir
kal
lt" s
poke
n by
the
hear
ing
subj
ect H
3.
13.,=
2Z'a
*L.:3
7,1
sec.
s u:
Dl
M E
d
a
sw';'
tV4,
0*-1
g./X
.."4"
.
f-=
OT
-4
segm
ents
no:
Woe
Jois
eim
nsie
ntIc
clus
ive
:Hoo
t lye
.ote
rat
Joso
lfo
wel
like
"ron
sitio
nal
-
;41.
;,
01t,!
101:
3# :'
ihri
Iitt
i::41
";:4
":-7
"- A
l.
123ei
ttlin
tifilt
illia
ikiii
ill
4
it,iii
:iiiii
ii'11
%-
56
11iii
iiiili
tlitiV
i
78
912
13
-..1
:11.
,'`i
it
15
.ii;
ithal
ialli
llitit
ilii
16
. 17
1111
MT
li:til
i:1
11IL
UI1it
i -....
.i.r
_
25
iditt
ilial
lialli
iillii
iiiiii
niiii
iiiiii
ii,_
26
,.1 t3
:., 27
28
,=
., ...
.-
-_
29
_Ild
liiill
ifil:
:1;;1
1:0'
;01!
!iri
lilill
tiniii
iiiiii
iiai.;
iii;,i
liiki
liniii
iiii
32
.,..
.,....
...itk
iiiill
halt4
iliiii
33
i,
.-.. 4
,
36
.1i
4r.
__ ..
..-
.i
38
..
-111
1111
11ri
_IP
-
--'-
ii
ii
II
--*
0.1
2.3.
4 .5
.6.7
.8.9
1.01
1.1
12 t3
1.4
1.5
1.6
1.7
1.8
1.9
2.0
.1 2
2.
..
..
..
..
..
..
..
sec.
Fig.
III
-B-2
. Son
agra
m a
nd f
eatu
re p
atte
rn o
f th
e te
st s
ente
nce
"sol
ensk
iver
men
det
firka
llt"
spok
en b
y th
e de
al s
ubje
ct D
I.
D2
su=
lE
nf
isn
Er
tt m
End
al
t
0-1;
i.
Seg
men
ts n
c
Voi
ceN
oise
Tra
nsie
ntO
cclu
sive
Fric
ativ
eLa
tera
l.N
asal
.V
owel
like
Tra
nsiti
onal
1
,. T
.: ..,
ildni
liii
2
, ii... 4
tmiii
ngoi
llidi
ki,.
678
9
,Tii:
iii-i
i11
1111
1114
1ton
10
,t
11
1-4,
..11
.. ..
18
lidi
'0":
41ili
ltlip
1
19
..,.0
1.1.
4.
20
:::
ir:...
.1,
.. ..i,..
...4
.i4lia
....
26
.;;1.
i!ii
,.i.,i
p.%
.!-
27
;...
it:.
.ii;
:..-
....
hi'
,ip.i,
ili.
:..,t
,iiii
ilill
32
,I ii'
34
.,..-
2831 .
Ia"
464
4^w
Aw
. . a
". A
a...
. I "
waf
ta.
. ""A
s
..d
W/.1
U.
IwJ
..
Fig.
III
-B
-3.
Son
agra
m a
nd f
eatu
re p
atte
rn o
f th
e te
st s
ente
nce
"sol
ensk
iner
men
det
ar
kallt
" sp
oken
by
the
deaf
sub
ject
D2.
1
Seg
men
ts n
o:1
810
1114
15 1
62
224
28 2
931
Voi
ceN
oise
Tra
nsie
ntO
cclu
sive
Fric
ativ
eLa
tera
lN
asal
Vow
etlik
eT
rans
ition
al
III
1111
1111
111
MIM
I11
1111
1111
1111
11
IIII
IIII
I
1111
111
1111
1111
1111
1111
1111
1111
11
111
1111
1111
1111
III
1111
1111
MIM
III
IE
MI
MIN
NM
IM
IMI
MIM
I
I111
1111
1111
11E
MI
III
III
1
1111
1111
1111
1lin
III
1111
1111
111
1111
111
ION
III
illill
I11
1111
1111
111
III
IIII
IIII
IH
MI
IIII
I11
1111
1111
1110
MI
III
I11
1111
1111
11 M
I II
1 1
Fig.
III
-B-4
. Son
agra
m a
ndfe
atur
e pa
ttern
of
the
test
sen
tenc
e "s
olen
skin
er m
en d
eta:
rka
llt"
spok
en b
y th
e de
af s
ubje
ct D
3.
= H
1aH
2-
H 3
x =
D 1
0 =
D 2 D 3
X
X X
X a 0U
0
A.
lEni
f 1
i:1 n
'Er
Iit
ImIE
In14
f cl)I
d e:
WW
I k1
al l
I It
a
A rip
X s. A
A 0
0X A I0
A U0
X 0
A
AO
Ix0 A
Fig.
III
-B-5
. Pho
nem
edu
ratio
ns in
the
test
sen
tenc
e"s
olen
ski
ner
men
det a
r ka
llt"
mea
sure
d at
the
thre
e he
arin
g su
bjec
ts H
1, H
2,H
3, a
nd a
t the
thre
e de
af s
ubje
cts
D1,
D2,
and
D3.