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,

,&ii;

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.