Applying the Vowel Formant Dispersion (VFD) method to the study of reduced or alterered vowel productions

Full text

n=1
Introduction
➡
Formant centre frequencies are used as acoustic measures of
vowel quality because they correlate with vowel articulation
➡
A number of measures based on vowel formants have been
derived to capture whole-system effect of vowel production.
➡
Most commonly used is Vowel Space Area
1
(VSA), the area of
the quadrilateral/triangle enclosing the vowels [a], [u], [i], [æ] or
[a], [u], [i] that constitute acoustic extremes as represented by
average formant frequencies.
➡
However, VSA may not be powerful enough to capture small
changes in vowel articulation due to effects of treatments for
patients with Parkinson’s disease (PD). AVS has not been
evaluated in this respect.
➡
Alternative measures , e.g. Vowel Articulation Index
2
(VAI) or
its inverse FCR
3
, and Average Vowel Space
4
(AVS) are more
focused on detecting expansion or reductions in vowel spaces.
-
These metrics do not give detailed insight into the nature of
the change in vowel articulation.
➡
Further, these metrics reduce all obtained vowel formant
measurements into a single metric,
-
all vowel formant measurements are reduced to a single
metric, giving
-
a substantial loss of statistical power
-
a considerable reduction in quality of intra-speaker models of
articulatory proficiency.
Vowel Formant Dispersions (VFD) in Hz
500
1000
1500
Normal Positive treatment effect Negative treatment effect
Hypokinetic dysarthria
Different kinds of articulation
n=1
F
2
F
1
400
600
800
1000 1500 2000 2500
F
2
F
1
400
600
800
1000 1500 2000 2500
+
+
Relocation of the vowel space
w
i
=
tan
−1
F
1
i
− F
1
c
F
2
i
− F
2
c
⎛
⎝
⎜
⎞
⎠
⎟
for F
2
i
> F
2
c
tan
−1
F
1
i
− F
1
c
F
2
i
− F
2
c
⎛
⎝
⎜
⎞
⎠
⎟
+
π
for F
2
i
> F
2
c
and F
1
i
≥ F
1
c
tan
−1
F
1
i
− F
1
c
F
2
i
− F
2
c
⎛
⎝
⎜
⎞
⎠
⎟
−
π
for F
2
i
< F
2
c
and F
1
i
< F
1
c
π
2
for F
2
i
= F
2
c
and F
1
i
> F
1
c
−
π
2
for F
2
i
= F
2
c
and F
1
i
< F
1
c
0 for F
2
i
= F
2
c
and F
1
i
= F
1
c
⎧
⎨
⎪
⎪
⎪
⎪
⎪
⎪
⎪
⎪
⎩
⎪
⎪
⎪
⎪
⎪
⎪
⎪
⎪
F
2
c
=
1
n
F
2
i
i=1
n
∑
{V (F
2
, F
1
) : F
1
< F
1c
}
F
1
c
=
1
n
F
1
i
i=1
n
∑
d
i
= V
c
V
i
 
= (F
1
i
− F
1
c
)
2
+ (F
2
i
− F
2
c
)
2
Math corner
d
c
= V
c
1
V
c
2
 
= (F
1
c
1
− F
1
c
2
)
2
+ (F
2
c
1
− F
2
c
2
)
2
Fredrik Karlsson & Jan van Doorn
Umeå University, Sweden
F
2
F
1
200
400
600
800
500 1000 1500 2000 2500
n=n
Poster presented at the 14th Meeting of the International Clinical Phonetics and Linguistics Association, in Cork, Ireland, June 27–30 2012
F
2
F
1
200
400
600
800
500 1000 1500 2000 2500
Vowel Formant Dispersion (VFD) in Hz
200
400
600
800
1000
1200
[u]-corner [i]-corner [ae]-corner [a]-corner
Vowel Formant Dispersion
A new metric, Vowel Formant Dispersion
5
(VFD) is proposed to
address these issues for clinical groups, to provide
-
more detailed interpretations of speakers’ production proficiency
-
increased reliability of within-speaker estimates of this
proficiency.
A vowel space centre is defined for each speaker as the point placed
at the mean F
1
of all vowels in the system, and the F
2
of those vowels
that have F1< F
1
(center line) i.e. vowels that are more closed.
-
This definition aims at supporting a robust placement of the
center across vowel spaces with differing properties
Each vowel is subsequently viewed as a vector originating from the
center point, each having a length (the VFD length) and an angle
component.
VFD applied to studies of clinical groups
➡
In statistical group comparisons, all measured vowel productions may be
included in the model
➡
A robust placement of vowel space center affords inferences concerning the
direction of changes in vowel articulation range to be drawn.
➡
The VFD lengths further affords interpretations concerning changes in the
variability of speakers in realizing targets.
➡
By investigating the placement of the vowel space center, a descriptive
analysis of whole-system relocation, that may be a desirable treatment
outcome in transexual patients, is also possible.
-
However, it should be kept in mind that this is a point estimate of the
relocation, and therefore not robust to the effect of individual variation.
References
1. Kent, R. D., & Kim, Y.-J. (2003). Toward an acoustic typology of motor speech disorders. Clinical Linguistics & Phonetics, Clinical Linguistics & Phonetics, 17(6), 427–445.
2. Roy, N., Nissen, S. L., Dromey, C., & Sapir, S. (2009). Articulatory changes in muscle tension dysphonia: evidence of vowel space expansion following manual circumlaryngeal therapy Journal of
communication disorders, Journal of communication disorders, 42(2), 124–135.
3. Sapir, S., Ramig, L. O., Spielman, J. L., & Fox, C. (2010). Formant Centralization Ratio: A Proposal for a New Acoustic Measure of Dysarthric Speech. Journal of Speech, Language, and Hearing
Research, 53(1), 114–114.
4. Lane, H., Matthies, M. L., Perkell, J. S., Vick, J., & Zandipour, M. (2001). The Effects of Changes in Hearing Status in Cochlear Implant Users on the Acoustic Vowel Space and CV
Coarticulation. Journal of Speech, Language, and Hearing Research, 44(3), 552–563.
5. Karlsson, F., & van Doorn, J. (In review) Vowel formant dispersion as a measure of articulation proficiency, The Journal of the Acoustical Society of America.
F
2
F
1
200
400
600
800
500 1000 1500 2000 2500
Te
A
Statistics
Applying the Vowel Formant Dispersion (VFD) method to the study of reduced
or alterered vowel productions
måndag 24 september 12