the acoustics of nasals and laterals...24.963. linguistic phonetics. the acoustics of nasals and...
TRANSCRIPT
24.963�
Linguistic Phonetics�The acoustics of nasals and
laterals
1
0
1
2
3
4
5
FREQ
(kH
z)
0 100 200 300 400 500 600 700 800 900
Image by MIT OCW. Adapted from Stevens, K. N. Acoustic Phonetics.
Cambridge, MA: MIT Press, 1999, chapter 6.
Assignments • Talk to me about a paper topic, this week or next.
2
Damping and bandwidth
• When an input of energy sets a body vibrating, the resulting vibrations die out as energy is dissipated through friction, transmission of energy to surrounding air, etc.
• A wave of this type is described as damped.
3
0 .001 .002 .003 .004 .006 .007 .008 .009.005 .01Time in seconds
Image by MIT OCW.Adapted from Ladefoged, Peter. Elements of Acoustic Phonetics.Chicago, IL: Univeristy of Chicago Press, 1962.
Damping and bandwidth
• The more heavily damped a sinusoid is, the wider the bandwidth of its spectrum.
• A wider bandwidth implies a lower spectral peak, other things being equal, since the energy in the wave is distributed over a wider range of frequencies.
4
Air
pres
sure
Am
plitu
deA
mpl
itude
Am
plitu
de
Air
pres
sure
Air
pres
sure
0 .01 0 500 1000Time in seconds Frequency in Hz
Various waves and their spectra.004 .005.003
Time in seconds.002.001
Complexwave
1,100 Hz
1,000 Hz
900 Hz
0
a b c d e
Image by MIT OCW. Adapted from Ladefoged, Peter. Elements of Acoustic Image by MIT OCW.
Phonetics. Chicago, IL: Univeristy of Chicago Press, 1962. Adapted from Ladefoged, Peter. Elements of Acoustic Phonetics. Chicago, IL: Univeristy of Chicago Press, 1962.
Damping and bandwidth
• Standing waves in the vocal tract are damped due to friction, radiation losses to the outside air, absorption of energy by the elastic vocal tract walls etc.
• Greater surface area results in more damping (and broader formant bandwidths).
• Coupling the nasal cavity to the vocal tract in in nasal sounds increases the surface area of the vocal tract.
5
Nasal stops
• A uvular nasal can be modeled in terms of a tube which is closed at the glottis and open at the nostrils.
• It is not a uniform tube. • The effective length of the tube is greater than the length
from glottis to lips – 21.5 cm (Johnson) or 20 cm (Stevens) vs. 16 cm.
• So the resonances of the tube are at lower frequencies (and closer together) than in a vowel.
6
Nasal stops
• In nasal consonants formed with an oral constriction further forward than the uvula, the oral cavity forms a side branch on the pharyngeal-nasal tube.
• This side branch is longest in labials, shortest in velars.
7
(A) (B)
(m)
Nose
Mouth
Pharynx
(A) An X-ray tracing of the positions of the articulators during (m).(B) A tube model of this vocal tract configuration.
Image by MIT OCW.Adapted from Johnson, Keith. Acoustic and AuditoryPhonetics. Malden, MA: Blackwell Publishers, 1997.
Nasal stops • The side branch is a tube open at one end and resonates
accordingly. • The side branch is not coupled to the outside air. Its
resonances are zeros in the signal radiated from the nose -i.e. those frequencies are attenuated.
• Zeroes also reduce the amplitude of higher frequencies, so nasals are characterized by less high frequency energy than vowels.
• Frequencies of the zeroes depend on the size of the side branch, e.g. 8 cm for [m], 5.5 cm for [n].
8
Nose
Mouth
Pharynx
A tube model of this vocal tract configuration.Image by MIT OCW.
Adapted from Johnson, Keith. Acoustic and AuditoryPhonetics. Malden, MA: Blackwell Publishers, 1997.
5000
Nasal stops
5000
Fre
qu
ency
(H
z)
Fre
qu
ency
(H
z)
0 0 0.03722 0.2246 9.767 9.988
Time (s) Time (s)
[m] my[n] nigh
9
Cues to place in nasal stops
• Formant transitions – as in oral stops• Nasal murmur differs according to place of articulation,
but in itself is a weak cue – murmur is relatively quiet.• But the interval around the release of the nasal may
provide significant place cues:– Nasal place is identified more accurately from 6
periods centered around release compared to 6 periodsof the murmur or formant transitions (Kurowski &Blumstein 1984)
– But the differences are small: 12% vs. 13% vs. 20%• The interval around the onset of nasal closure does not
provide comparable cues to place (Repp & Svastikula 1988)
10
Laterals
• Laterals also have a side branch - the pocket behind thetongue tip is a side branch to the main tube(s) passingaround the side(s) of the tongue.
11
(a)
[l]
13.5 cm
2.5 cm
2.5 cm40 cm3
A = 0.17 cm2
l = 1.0 cm
Image by MIT OCW.
Image by MIT OCW.Adapted from Johnson, Keith. Acoustic and Auditory Phonetics Malden, MA: Blackwell Publishers, 1997.
Laterals
• Laterals also have a side branch - the pocket behind the tongue tip is a side branch to the main tube(s) passing around the side(s) of the tongue.
• Laterals are thus also characterized by zeroes - the lowest appears between F2 and F3, often significantly reducing the amplitude of F2.
• The presence of zeroes and the coronal constriction reduce the intensity of laterals compared to most vowels.
• On spectrograms, laterals look similar to nasals, but differ in the location of formants and zeros, and in their effects on neighboring vowels.
12
Laterals
Fre
qu
ency
(H
z)
0
5000
Fre
qu
ency
(H
z)
0.03722 0.2246
0
5000
7.507 7.726
Time (s)Time (s)
[n] nigh [l] lie
13
Laterals
5000
Fre
quen
cy (
Hz)
0 30.39 30.67
Time (s)
[dli(ma)] (Hebrew)
14
Nasalized vowels
• Nasalized vowels involve a complex vocal tractconfiguration.
15
Nose
Mouth
Pharynx
Image by MIT OCW.Adapted from Johnson, Keith. Acoustic and AuditoryPhonetics. Malden, MA: Blackwell Publishers, 1997.
Nasalized vowels
• The coupled nasal cavity contributes both formants andanti-formants (poles and zeroes) which combine with theformants of the oral tract.
• The net acoustic effect of velum lowering depends on thelocations of the formants in the corresponding oral vowel(Maeda 1993)
16
17
Nasalized vowels
Z2M
agni
tude
(dB
)Z1
NF1
A
NF2
F2
F1F1
-30-20-10
10203040
0
Z2
Frequency (kHz) Frequency (kHz)
Z1
NF1
NF2
F2F1
F3
F3
543210 543210
0
Frequency (kHz)
Mag
nitu
de (d
B)
Z1
NF1
C
F1
F1F2
-30-20-10
10203040
0
5432Frequency (kHz)
Mag
nitu
de (d
B)
Z1
Z2
NF2
EF1
F2F3
F2F2F2'
F3
10
-40-30-20
102030
-100
Z2
Frequency (kHz)
Z1
MF1
D
NF2
NF1F2F1
F11 F2 F3F3
543210 543210
0
B
OralNasal (NC = 0.4) Oral
Nasal (NC = 0.4)
OralNasal (NC = 0.2)
OralNasal (NC = 0.4)
OralNasal (NC = 1.6)
i
ε
ο
u
''
''
'
'''
''
Image by MIT OCW.Adapted from Maeda, Shinji. "Acoustics of vowel nasalization and articulatory shifts in French nasal vowels."In Nasals,Nasalization and the Velum. Edited by M. Huffman and R. Krakow. New York, NY: Academic Press, 2003, pp. 147-170.
Nasalized vowels
• As a result it is difficult to give a general acousticcharacterization of nasalization on vowels.
• Maeda: nasalized vowels generally have a flat, lowintensity spectrum in the low frequency region.
– Addition of the lowest nasal formant in combinationwith broad bandwidth F1 can create a broad, lowfrequency prominence.
18
0 100 200 300 400 500 600 700 800 9000
1
2
3
4
5
FREQ
(kH
z)
0 100 200 300 400 500 600 700 800 900
'Don's bed''dot bend'
Image by MIT OCW.Adapted from Stevens, K. N. Acoustic Phonetics. Cambridge, MA: MIT Press, 1999.
Nasalized vowels
• Chen (1997) proposes the measures A1-P0 and A1-P1 toquantify nasalization.
– A1 is amplitude of F1– P0 is amplitude of the low frequency nasal peak.– P1 is amplitude of a higher nasal peak, between F1 and F2.
19
• A1 amplitude of F1• P0 amplitude of lower frequency
nasal peak - intensity of harmonicwith greatest amplitude at lowfrequency.
• P1 amplitude of nasal peak -measured as amplitude of highestharmonic near to 950 Hz.
• These measures are sensitive tovowel quality - Chen proposescorrections to adjust for this.
20
543210
0
40
20
P1
FREQ (kHz)
MA
G (d
B)
P0A1
60
Non-nasal Vowel
543210
0
40
20
P1
MA
G (d
B)
P0 A160
Nasal Vowel
FREQ (kHz)
Image by MIT OCW.Adapted from Chen, Marilyn Y. "Acoustic correlates ofEnglish and French nasalized vowels." The Journal ofthe Acoustical Society of America 102 (1997): 2360.
Nasalized vowels
• Perceptually, nasal vowels are less distinct from eachother than oral vowels, both in terms of confusability(Bond 1975) and similarity judgements (Wright 1986).
21
Dimension II
The first two dimensions in theperceptual space as determined byINDSCAL
F1 versus F2 plot of the mel values ofthe stimuli
F2 in mels
-.4
-.2
0
.2
.4
Dim
ensi
on I
-.6 -.4 -.2 0 .2
u
υυ
ae
a
a
e ~e
~~o
~
o
ε~ε
~ε~
u
ιι
400
600
800
1600 1400 1200 1000
F 1 in
mel
s
u
ae
a
e~e
~
~o
o
ε
u
ι
ae
~ae
υ
υ
a
~ι
Image by MIT OCW.Adapted from.Wright, James T. "The behavior of nasalized vowels in the perceptual vowel space." In ExperimentalPhonology. Edited by J. J. Ohala and J. J. Jaeger. Orlando, FL: Academic Press, 1986, 45-67.
17
Nasalized vowels
• More specifically, high nasalized vowels are perceived aslower than their oral counterparts and low nasalizedvowels are perceived as higher than their oralcounterparts.
• Wright speculates that this may be related to the fact thatFN1 is below F1 in low vowels and above it in high andmid vowels in his stimuli.
Nasalized vowels
• This greater confusability is reflected in the fact that, inlanguages with contrastive vowel nasalization, the nasalvowel inventory is always the same as or smaller than theoral vowel inventory, never larger (Ferguson 1963, Ruhlen1973).
23
Beembe (Congo) Dakota (Yankton-Teton)
~ ~ ~ ~i u i u i u i u~ ~e o e o e o
~a a ~a a
Portuguese (Continental) Ebrie (Niger-Congo)
~ ~i u i u i u~ ~e o e o ~ ~e o e o
ε c
~a a~a a
Image by MIT OCW.
pin-pen neutralization
• Southern US (and Bakersfield) accents:– pin-pen, him-hem are homophonous, pit-pet contrast.
• Vowels are substantially nasalized before nasals in English.• Nasalization reduces the distinctiveness of F1 contrasts• Oral [I]-[E] are sufficiently distinct, nasalized [I) ] are]-[E)
not.
24
Phonological map of the US removed due to copyright restrictions.Source: http://www.ling.upenn.edu/phono_atlas/home.html.
MIT OpenCourseWarehttps://ocw.mit.edu
24.915 / 24.963 Linguistic PhoneticsFall 2015
For information about citing these materials or our Terms of Use, visit: https://ocw.mit.edu/terms.