All images in this file are courtesy of MIT Press. Used with permission.
: 6.541J Handout
L
2
uAI Ar
I
/P_
Figure 9.34 Midsagittal sections of the vocal tract for the liquid consonants /r/ (left) and /I/
(right). The midsagittal sections do not reflect the fact that in each case there is an acoustic
path around the lateral edges of the tongue blade. The representation of /1/ is adapted from
Narayanan et al. (1995b).
I
I
aw
a:
u-
U.
Figure 9.35 Spectrograms of the utterances /do'rod/ (left) and /d'lod/ (right) produced by a
male speaker. The time course of the first-formant frequency during the liquid consonant is
overlaid on the spectrogram.
(a)
8 cm3
Mc
(b) Z
b
RC
M
Figure 9.36 (a) Schematized model of the vocal tract used for estimating the acoustic characteristics of a retroflex consonant. The side branch representing the space under the tongue is
shown, together with the constrictions formed by the tongue blade and the lips. (b) Equivalent
circuit used to obtain the approximate frequency of the front-cavity resonance of the configuration in (a). The volume of the front cavity is represented by Cf, and the constrictions posterior and anterior to the cavity are represented by R, M, R,, and M,. Zb is the impedance of
the cavity behind the constriction.
-c~----
r
207-
I
20
o
1
VA
V
2
V'
3
FREQ (kHz)
-
4
5
FREO(kHz)
A N
V
V
V
LSA
V
._
..
ss .-
VV
VV
~
V"
/r
.
V
-
173ms
Figure 9.40 Spectra sampled at several points in the utterance /da'rod/, for which a spectro­
gram is shown in figure 935. Spectra 1, 2, and 3 are calculated using a short (6.4 ms) time win­
dow, whereas spectrum 4 uses a longer (30 ms)time window. Spectra 2 and 4 are sampled near
the most constricted part of the consonant /r/, and show the dose proximity of F2 and FR, at
about 950 and 1350 Hz, with evidence of a weak F3 peak at 2300 Hz and a zero just below 2000
Hz Spectra I and 3 are sampled during the initial movement from the initial vowel and the final
movement toward the final vowel, respectively. These spectra show a double peak correspond­
ing to FR and F3 in the frequency range 1.5 to 2.5 kHz.
N
pi,
_r
LL
a:
LU
a
U
ir5
I
��--�---
LL
-�
-------
---__�_�_i···li�n�·�I·n;
Table 9.2 Ranges of average values of frequencies (in hertz) of first three resonances of
prevocalic /r/ as reported by Espy-Wilson (1987), Hagiwara (1995), and Westbury (1995)
Female speakers
Male speakers
FI
F2
FR
360-480
1030-1240
1800-2050
330-430
880-1200
1380-1610
Note. The different studies used different numbers of speakers and contexts.
-
13.5
cm
40cm
C
A=0.17 cm2
/ I1.0 cm
Figure 9.43 Schematization of the vocal tract shape for purposes of calculating the acoustic
characteristics of a lateral consonant. Approximate dimensions are shown for the back cavity, the
side branch formed by the tongue blade, and the cavity in front of the constriction. The dashed
line indicates the midline of the acoustic path from glottis to lips.
n­
TV
)
.
.
_
I
N
I
1.5
-
>­
I
Z
I
1.0
'I
Fl
0
LLJ
_
V
,,r
ms)-
50
I
I, -
100
~
150
n0
TIME (ms)
Figure 9.48 Left, Spectrogram of the utterance a le. Right, Frequencies of first two formants
preceding and following the release of /1/, with points given at each glottal period.
c
40
o
5
4
3
FREQ kHz)
2
1
0
216ms
131ms
­
60
4040
,,
r
5
4
3
FREO (kHz)
2
1
0
v
V-V
VV,
'
i
iI
60
40
0
20
V
"\_
:
III'
i!XI
.
___7
-7 -7
-3-_
L
_
1
HF
I .,-
---- 7
�1 7
157s
I
I
-~.-..v-~,V'
V
_
_
i1i
fh
I
_
_
2
V-
_I
._-
A
t-=
·. ·
7
I
FREU (KHz)
~-X
, v ,
199ms
,- ~­v
I5
Iv
195ms
Figure 9.46 Spectra sampled at several points within the utterance /da'lod/, a spectrogram of
which is shown in figure 935. Spectra I to 5 are obtained with a time window of 6.4 ms, cen­
tered on individual glottal pulses. At time 157 ms (spectrum 2), there is evidence of a double
peak in the F3 range, apparently resulting from two slightly asymmetrical acoustic paths around
the tongue blade. The increase in spectrum amplitude of the F4 peak (around 3400 Hz) is espe­
cially abrupt from 217 to 233 ms (spectra 4 and 5). Spectrum 6 is obtained with a 30-ms time
window centered at a time immediately prior to the release (about the same time as spectrum 3).
The peak at 1400 Hz is assumed to be a tracheal resonance.
--------------�--··--------·------�··--------
--------
Wose Aue to tLrobtIe@e it AOll
AZL/c_ (a)
-j -_F-------
(b)
~~i~~~P~
(c)
Figure 2.31 Airflow impinges on the wall of a tube downstream from a constriction, to gen­
erate turbulence noise at the wall.
Figure 2.30 Schematic representation of three types of acoustic sources due to turbulence in
the airstream near a constriction: (a) no obstacle; (b) obstacle in the airstream downstream from
the constriction; (c) fluctuation in flow through the constriction.
pQ/A
cavities upstream
cavities downstream
from source
from source
p1/A Ps,
_ +
Figure 2.32 Equivalent circuit representation of a turbulence noise source as a one-dimensional
dipole with two equal and opposite volume velocity sources Q (top) and as a sound pressure
source (bottom).
Ag
Ac
Figure 2.36 Tube with two constrictions, with turbulence noise generated at both con­
strictions. The cross-sectional areas of the constrictions are A, and A, as indicated, representing
the glottal and supraglottal constrictions.
�I_�
�
N
E
I
I
I
I
'
I
I
I
w 40
420 cr
3
/s
(o)
CD
o
0
- I IbU
20
-20
015
c
u)
I
I
I
2
5
FREQUENCY (kHz)
I
I
I
I
I
I
I
I
I
.
I
,
,
I
_
..
E
20
420 cm/s
-U
0
0
=
_
o,
~
=
(b)
--
Ib
--
oo
0
N
4O
20
I­
I1
I
0.5
1
I
I
2
I
I
I
5
10
FREQUENCY (kHz)
Figure 2.33 Spectrum of sound pressure source p, for a configuration similar to that in figure
2.30b, for two values of airflow. Distance from the constriction to the obstacle is 3 cm, and the
diameter of the (circular) constriction is 0.32 cm. The cross-sectional area of the tube is 5.0 cm2 .
The spectra are in 300-Hz bands of frequency. (a) The ordinate is the absolute level of the sound
2
pressure source, in decibels re 1 dyne/cm . (b) The same curves are given, but with an ordinate
that is in units of volume velocity, to permit comparison with other volume velocity sources in
the vocal tract. The sound pressure source p, is normalized by dividing by the characteristic
3
impedance pc/5. For 0 dB on the ordinate, 5p5 /(pc) = 1 cm /s. When the area At of the tube
2
curves in (b) should be scaled up
from
5
cm
,
the
downstream from the constriction is different
3
by 20log (At/5) dB. For the upper curve in both panels, the value of 201og U A-25 = 212 dB,
where U =volume velocity in cubic centimeters per second and. A = area of constriction in
- 25
is different from 212 dB, the upper curve should be
square centimeters. When 20 log UA
5
3
212
dB.
Curves are based on experimental data of Shadle (1985a).
scaled up by 20 log U A-2
(a)
P.
(b)
Figure 2.42 (a) Equivalent circuit of the vocal tract that can be used to estimate the time
course of the turbulence noise source at the time-varying supraglottal constriction following the
release of a stop consonant. (b) Equivalent circuit containing the principal elements that deter­
mine the turbulence noise source.
-P 20
> J
J -Source
glottis
i
0
LiJJ
Iction
t
I
/
0
0.1
w 0
1
suprcaottl
S r3
/ /
>
at
Ag 0.2 cmZ
0.2
0.4
0.3
0.5
a 20
_j
--
et
awuree
>
supraglottOI
10 Constriction
>
-
source at
glottis
,
J I
0O
0.1
02
0.3
0.4
0.5
AREA OF SUPRAGLOTTAL CONSTRICTION AC (cm)
Figure 2.37 Relative level of noise source at each constriction of figure 236, as a function of
the cross-sectional area of the supraglottal constriction. The cross-sectional area A, is fixed at a
different value for each panel The constant K in equation (2.21) is assumed to be the same for
the noise generated at each constriction.
yMi
Icm
]cm
Figure 2.46 Schematic representation of the shape of the lips (top) and tongue surface
(bottom) as a labial or a velar stop consonant is released. At the instant of release, the length of
the constriction is relatively short, due to the distortion of the surfaces by the action of the
intraoral pressure.
-----------------
---~~~~~~~~~~~~~~~~---L~~~~
sTrp
e-Dy\640ho-ttt
-J
E 1.0
uw
Wo,
o<
200
-j
I
-100
0
100
i-I. O
E
C)
z
W
M
W
0
_
/I'
8-1.
D--200
-100
0
TIME FROM RELEASE (ms)
Figure 7.1 Schematic representation of articulatory movements when a stop consonant is pro­
duced in intervocalic position. Time from release of the closure is plotted on the abscissa. The
upper panel is a representation of the cross-sectional area at a point in the vocal tract where the
constriction is formed. The lower panel indicates the type of trajectory that might be followed
by a secondary articulatory structure such as the mandible or the tongue body as the inter­
vocalic stop consonant is produced. The durations, rates, and extents of movement are selected
to be within the ranges observed in actual utterances, but do not necessarily correspond to a
particular stop-consonant trajectory. The time course of the cross-sectional area may be modified
somewhat from the smooth trajectory shown above as a consequence of forces due to the
intraoral pressure.
.
Figure 7.2 Midsagittal sections through the vocal tract during the production of stop con­
sonants produced by dosing the lips (left), raising the tongue tip to the alveolar ridge (middle),
and raising the tongue body (right). Adult male speaker. (From Perkell 1969.)
ii
4
PS
Ps
(o)
(b)
Figure 7.3 (a) Low-frequency equivalent circuit of the vocal tract when there is a complete
closure in the oral cavity. (b) Approximate equivalent circuit that is used for calculating intraoral
pressure P. and glottal airflow Us immediately following closure of a stop consonant. See text.
--·----------_
----
^-·CI-*-��
_I__·-r��-·--
�'r�
�_____
-
_*�---------p--P-~"~
I
E
IFpolate
0
0
-I
w
I
dorsum of
tongue
w
TIME
0
FROM
RELEASE
2
d
-J
(ms)
Icm
(0)
3
I
0
E
4
C-)
0
1
' ®
(b)
-j
0
TIME
N
-
E
00 8000
FROM
RELEASE
I
(ms)
T-20-I0
O
0 20 30
TIME (sec)
I
INTRAORAL
Figure 7.6 (a) Schematic representation of the surfaces of the tongue and the palate at several
times during the release of a stop consonant produced by making a dosure with the tongue
body. The dashed lines indicate the tongue contour if there were no intraoral pressure, and the
solid lines show the contour as modified by forces due to the intraoral pressure. (b) Estimated
cross-sectional area vs. time at the constriction without the influence of intraoral pressure (dashed
line) and with intraoral pressure (solid line). The labeled arrows indicate the points in time corresponding to the panels in (a).
PRESSURE
>'6000
Ad4000
VELAR
LABIAL
(n 2000
to
0
c
a
0
10
TIME
20
30
40
50
FROM RELEASE (ms)
Figure 7.8 Calculated airflows (top two panels) and intraoral pressures (bottom panel) as a
function of time following the release of voiceless unaspirated stop consonants. The constriction
area follows a trajectory of the form A, = A,,(
- e-'), where A. is taken to be 1.0 cm2 ,
-x
and /r = 100 s-l for labials releases and 25 s for velars. The area for the velar release is
further modified as in figure 7.6. The glottal area is assumed to be constant at 0.1 cmz. Initial
intraoral pressure is 8 cm H20, and other parameters are as given in figure 7.4. Negative U,
indicates that the walls of the vocal tract displace inward following the release. Note that
u =
-�-
+ u.
---�------�--�
�-----
10-­
FFRICATIC IN
ASPIRATION
5
VOICING
eIrIT
TIPAM
I
' ', 'ii
4
w
!ri iiii
I
I
3
2:
0
'I
oU
LI
I
I
I.
A1
40
20
0
A
vv7
vy
#'
-1
AA
TIME (ms)
0
0
100
200
Figure 7.17 Schematic representation of sequence of events at the release of a voiceless unas­
pirated stop consonant. (From K. N. Stevens, 1993b.)
2
C,
6O
2~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Ii
13
(9
4
2
0
1
2
r
V
3
FREO (kHz)
GAl
4
5
AA0h~h
kIK2Y
Yp'
ii
ii~lkiv
IIill
VIv
Figure 7.20 Acoustic data obtained from an utterance of a voiceless unaspirated stop con­
sonant /p/ in the intervocalic environment /apa/ (the same utterance as that displayed in figure
7.12a). A spectrogram of the utterance is shown at the top, and spectra and waveforms at the
burst and 18 ms later are given below. The spectra are discrete Fourier transforms calculated
with a Hamming window of 6.4-ms duration.
------
-- �-----�-11�1
�
-��
--
~"
�"""~"�··-
t
-
_
/a E
L,
N
I
U:
60
0
1
2
3
FREQ (kHz)
4
5
3sms
e
ts
E
a
4
E
0
t
2
3
FREO (kHz)
113ms
4
5
C
0
Y
1
_
\
2
/
\
3
FREQ kHz)
I
­
4
307ms
Figure 7.27 Acoustic data obtained from an utterance of a voiceless unaspirated stop con­
sonant I/t/ in the intervocalic environment /at/. The four spectra below the spectrogram are
sampled at selected points before the consonant closure and after the release, as indicated by the
waveforms and time windows below the spectra See legend for figure 720.
5
-
�
3D
'
'
-
'
.
F3
F3
30__
____
25
I
z
a
o
z
F2
O5
15
tJ
i
2.5
[odo]
2.0
[E de]
,I
LL
1.0
0.5-
A
05
I I
-8.0
I
I
-40
-
0
Fl
C
, I
^^
0
40
80
100 - -80 -40
0o
TIME FROM CLOSURE OR RELEASE (ms)
I
.
I
I
LLL..
40
U
Figure 7.23 Measured frequencies of the first three formants for the utterances /ada/ (left) and
/ede/ (right) produced by a male speaker. See legend for figure 7.15.
2.5
V~.F3
2.5
ta
x
I
2.0
C)2
z
o.5
258ms
F2
[ g
-
A.0
-
0.5
0.5
.
,,
120
UO
40
[EE
1.0
0
n
.
-A
I --I I
A
--I
1..
F
-
n
n
4
R
11A
TIME FROM CLOSURE OR RELEASE (ms)
s
C
3
O
I
2
3
FREG (kHz)
4
5
Figure 7.31 Measured frequendes of the first three formants for the utterances /cga/ (left) and
/ege/ (right) produced by a male speaker. See legend for figure 7.15. The frequency of the major
spectral peak in the vicinity of F2 for /ege/ shows a discontinuity at about 40 ms. This dis­
continuity is presumably a result of interference by the third subglottal resonance for this utterance. The F2 curve is an estimate of the F2 movement if this interference were not present.
281ms
Figure 7.35 Acoustic data obtained from an utterance of a voiceless unaspirated stop con­
sonant /k/l in the intervocalic environment /ako/. See legend for figure 7.20.
��- ----- �---�
- 1-3 wi
21
CO.
-
ll---
-
tI
r.
r
E1
COO OW
tI
50
Wu
()
I
1 4=
-LAJ1LL2W
-------1
- l Z
-
I
i
TE
I,
II
I
~,FImI.Z
.
-
::Z- -
, I
-
:Nl - - - - -- =
-
o-
.-
I
;
i
.W
I
~
I/­ Iff-Z
wons
rod
7A
:sl
nlz
tl
_
I
�J_11 .�-.
I-I.
1
i
-­
7--.-"
- - - /-' 2
, , f
,
-7
iri
I_ ~.
I
IF -
I
n
I I ,:,:2H,^
1 Z
i L 1111 i
ii
~ii
Y7z1
-I- I I I
0-
r
3-
...
ga,.c
-
.-- I ,
t
'
I~~~~~
*.. **4.ral
ii /r
A i
-'I---
I I i
I I
7 ,
I I
< -_.4
I
I
-
4- A
'
' I
'
_
4-i
I
S __
I
7- '
-n
mC
.
!T-h
I
. I I
A
-=-
I C-1L
~_,
1
ruz
I
H
;H 11
I
3 =-
I
}
I__I__
S,, I,, I I 11r
­
-- 7-
1 it,
4
Z
­ill
,
Q~l
VI.
l
_
.
-
m
-
I I I I
aWrfAr
- vvtW'
,
W-
1r I
'
5- ­
{~{
b
,t~
I
I
r
--,'
__-
-- I
_
!-
Lv'v W''
I' - tI
1'#S_
I-'
{
LX,
*1-
Iw
_I
Ix
--
.IFi
6
IA-
I
'd=-
-Z
1Z~~s
Ir H
y
I VI U-40
1
. 1 .1 1 . II .
I
.
V
-.I-.
I
�l-�-�p
t
vy Y' mW Vy
IvREV'
I
yyv-
yv
i
-- �-�
I -
vy
vv x *
v
vy
r V,