Acoustic%20measurements%20on%20prosody%20using%20Praat

1
Acoustic measurements on prosody using Praat
Bert Remijsen Universiteit Leiden University of
Edinburgh
2
Overview
Brief motivation Introduction to Praat scripting
Measurement of gt Vowel quality gt
Fundamental frequency gt Voice quality and
intensity
3
Overview
Topics in relation to measurements gt Data
collection and processing gt How to measure it
in Praat gt (Semi-)automating measurements gt
Displaying the descriptive statistics gt
Inferential statistics
4
Motivation
5
Motivation
Why quantitative analysis of prosody? gt
quantitative results can be used to test
hypotheses
6
Motivation
Why quantitative analysis of prosody? gt humans
are bad at determining the acoustic cause of
prosodic variation by ear E.g. - controversy
on lexical stress - perception of
pitch-accent
7
Motivation
Why quantitative analysis of prosody? gt
Prosodic contrasts are often realized in terms
packages of prosodic correlates. E.g. stress
duration, vowel quality, intensity complementar
y quantity duration, vowel q. pitch-accent
fundamental frequency (f0), duration, etc.
8
Motivation
Why quantitative analysis of prosody with
Praat? gt Allows for measurement, manipulation,
and representation of the full range of
acoustic parameters. gt Relatively easy to
(semi-)automate procedures by means of scripts.
9
How to write a Praat script?
10
How to write a Praat script?
A. Try to start out from an existing script gt
For example, check on http//uk.groups.yahoo.com
/group/praat-users gt Praatscripts introduced
in this presentation can be found
at http//www.ling.ed.ac.uk/bert/praatscripts
11
How to write a Praat script?
B. Writing (part of) a script from scratch gt
Do the steps by hand for one item gt Display
them using Paste history gt Combine these steps
with control structures, guided by the manual.
12
How to write a Praat script?
An annotated script Script msr_duration.psc Fu
nction collecting durations for onset, nucleus
and coda of a target word, for each file in
list. Automatic.
13
How to write a Praat script?
Common components gt User interface (form
endform) gt Getting the input files (Read ) gt
Finding point of measurement (using TextGrid) gt
Measurements gt Writing output to file (e.g.
fappendinfo)
14
How to write a Praat script?
The dataset gt One long sound file e.g. the
whole recording session, with information on
sections in the TextGrid. gt One
item-per-file. If so, it is best to encode as
much useful information as possible in the
filename, in a structured way.
15
How to write a Praat script?

• gt One item-per-file. If so, it is best to
  encode as much useful information as possible in
  the filename, preferably fixed-width.
• E.g.
• dataset_code d2_2_012_s_1
• speaker_no
• item_no repetition_no
• s(ingular) / p(lural) SR

16
How to write a Praat script?
Reasons gt Saves work coding in statistics
package gt The fields in the name can be
searched with a Praat script (using string
pattern matching).
17
How to write a Praat script?
Script openlist.psc Function open specific
objects associated with item in list
18
How to write a Praat script?
Script openlist_specificitem.psc Function
This script searches on the item code the
third field in the name.
19
Measuring vowel quality
20
Vowel quality Measurement in Praat
How to measure formants in Praat? I. The
point of measurement II. An algorithm and a
protocol III. Semi-automating measurements
21
Vowel quality Measurement in Praat
I. The point of measurement possibilities gt
Where F1 reaches its maximum gt Small domain
centered on temporal mid point gt Averaged over
(middle x of) vowel.
22
Vowel quality Measurement in Praat

• II. An algorithm and a protocol
• Produce Formant object using default algorithm
  (Burg) and parameters (5 formants below 5000 Hz
  male / 5500 Hz female)
• Track using default values (male values female
  values 10 ).

23
Vowel quality Measurement in Praat

• 3. Protocol for when the value is incorrect
• E.g. weak F2 of high back vowels often missed
  F3 reported as F2
• Options - Use LPC with more coefficients
• - Retrack with changed F1/2 ref.
• The strategy is to be fixed within a single study.

24
Vowel quality Measurement in Praat

• III. Semi-automating the measurements
• gt Formant measurements should be checked. A
  fully-automated procedure is not an option.
• gt Instead automate all the repetitive actions.
  

25
Vowel quality Measurement in Praat
Script msrcheck_f1f2_indiv_interv.psc Function
Makes measurement as proposed above, Point of
measurement midpoint of an interval suitable
for analysis for monophthongs.
26
Vowel quality Measurement in Praat
Script msrcheck_f1f2_indiv_point.psc Function
Makes measurement as proposed above, Point of
measurement points on a point tier suitable
for analysis of di/triphthongs. gt These
scripts can easily be modified to process a
batch in one go still with check.
27
Vowel quality
Scaling
The formant values, once collected, can be scaled
in a number of ways 1. Individual frequencies
or frequency differences? gt Vowel height F1-F0
or F1 gt Advancement F2-F1 or F2
28
Vowel quality
Scaling
The formant values, once collected, can be scaled
in a number of ways 2. F1 x F2, or others
formants as well? gt F1 x F2 gt F0 x F1 x F2
x F3
29
Vowel quality
Scaling
The formant values, once collected, can be scaled
in a number of ways 3. Acoustic /
psycho-perceptual scale? gt hertz (Hz) gt
Logarithmic (ST) gt MEL gt ERB gt Bark
30
Vowel quality
Scaling
The formant values, once collected, can be scaled
in a number of ways 4. Cross-speaker
comparisons? gt z-transformation (Lobanov) gt
Gerstman gt Constant Log Interval Hypothesis
31
Vowel quality
Scaling
Ideal set-up for normalization (Adank 2003) gt
Individual frequencies rather than ?s gt hertz
(Hz) rather psycho-acoustic scale gt No need to
consider F0 and F3 gt between-speaker variation
z-transformation
32
Vowel quality Analysis / vowel plots
The formant values, can be interpreted best in a
vowel plot (F1 x F2). Characteristics of a good
vowel plot gt Inverted axes gt Over speakers
(so z-transformed) gt Categories labeled using
IPA
33
Vowel quality Analysis / vowel plots
The formant values, can be interpreted best in a
vowel plot (F1 x F2). Characteristics of a good
vowel plot gt Inverted axes gt Over speakers
(so z-transformed) gt Categories labeled using
IPA Praat can do it.
34
Vowel quality Analysis / vowel plots
Example - The vowels of Dinka /i,e,?,a,?,o,u/ -
Ellipses encircle 1 standard deviation - Separate
ellipses for compl. quantity - Values averaged
over 2 repetitions of 36 items uttered by 5
speakers.
35
Vowel quality Analysis / vowel plots
Example - The vowels of Dinka /i,e,?,a,?,o,u/ -
Ellipses encircle 1 st. dev. (68) - Separate
ellipses for compl. quantity - Values averaged
over 2 repetitions of 36 items uttered by 5
speakers.
36
Vowel quality Analysis / vowel plots

• Create a TableofReal, with, for each token
• gt praat-code for the IPA label (e.g. ? is
  \ep)
• gt z-transformed F1 and F2 sign inverted (I do
  this in SPSS)
• gt Header contains axis labels and no. of tokens

37
Vowel quality Analysis / vowel plots
Example formants_tor.txt File type
"ooTextFile" Object class "TableOfReal" numberO
fColumns 2 columnLabels "F2
(z-transformed)" "F1 (z-transformed)" numberOfRow
s 341 row 1 "iC" -1.6595 1.2794 row 2
"iC" -1.9973 1.2538 row 341 "oC" 0.6245
0.0380
38
Vowel quality Analysis / vowel plots
2. Open the TableofReal in Praat, and use
either gt Draw scatter plot to plot
individual values each token is marked by its
(IPA) label. or gt Draw sigma
ellipses ellipses, sized by user in terms of
st. devs. (sigma). (IPA) label plotted at
center.
39
Vowel quality Analysis / vowel plots
Either way, plot with no for Garnish and
Discriminant plane 3. In Picture window, add
marks on x and y axes, inverting the inverted
sign back to normal for example One mark
left... -2 no yes no 2 This gives a y-axis mark
in terms of z-scores of 2 at -2 on the y-axis,
without plotting -2.
40
Vowel quality Analysis / inferential tests
Characteristic inferential test ANOVA gt
within-subjects gt multivariate (dependents zF1
and zF2) gt factor(s) vowel quality (and e.g.
lexical stress / intonational accent / position
in phrase / etc.).
41
Measuring fundamental frequency
42
F0
Overview
gt Issues in measuring F0 gt Scaling gt
Descriptive stats
43
F0 Issues in
measuring F0
I. For detailed study about the realization of
tonal contrasts, consonants in target words
should be
nasals
liquids
approximants, rhotics
voiced fricatives
unvoiced fricatives, stops
44
F0 Issues in
measuring F0
BUT other may be more important such as the
availability of minimal-set data /ba1/ Low
level to remain /ba3/ High
level ancestor /ba121/ Rise-fall stiff /ba
12p/ Low Rise father /ba41/ Extra High
Fall to hit /ba21/ Low Fall to blow /
ba31/ High Fall when
45
F0 Issues in
measuring F0
II. F0 measurements need to be checked for
octave jumps etc. gt suggestion use a
semi-automated procedure
46
F0 Issues in
measuring F0
Script lst2f0check.psc Function This script
automates all the repetitive actions involved in
the checking of F0 tracks. It calculates the F0
track (Pitch object), plots it in the Picture
window, gives the opportunity to fix errors if
need be, and then writes the (fixed) Pitch object
to a file. Batch processing using file list.
47
F0 Issues in
measuring F0

• III. The point of measurement turning points
  can be determined
• gt by eye
• gt using mathematical modelling. MOMEL (Hirst
  Espesser) is implemented in Praat. See also
  recent work by Grabe Kochanski.

48
F0 Issues in
measuring F0

• Script momel_modif.psc
• Function Praat implementation of the MOMEL
  algorithm. (Original implementation in the MES
  signal processing package)

49
F0
Scaling
From physical F0 trace to psycho-acoustic
track. 1. Normalization for the logarithmic
nature of pitch perception gt hertz (Hz) gt
semitone (ST) gt Equivalent Rectangular
Bandwidth (ERB)
50
F0
Scaling
From physical f0-track to psycho-acoustic
track. 1. Normalization for the logarithmic
nature of pitch perception gt hertz (Hz) gt
semitone (ST) gt Equivalent Rectangular
Bandwidth (ERB) Latest news semitone is best
(Nolan 2003).
51
F0
Scaling
2. Normalization across speakers gt No need
to normalize for slope differences expressed in
ERB or ST. gt Absolute values can be
normalized using the z-transformation.
52
F0 Analysis / Plotting
tracks
How to interpret the data, and communicate
tendencies to others? The problem gt Averages
of F0 measures expressed as numbers in Hz are
hard to interpret. ST, ERB and z- scores are even
harder to interpret. gt Visual illustration by
means of F0 tracks of individual cases fail to
exploit the dataset.
53
F0 Analysis / Plotting
tracks
The solution gt Represent F0 visually across
speakers, by means of tracks normalized for
time. gt I.e. graph used as a descriptive stat
(reports average)
54
F0 Analysis / Plotting
tracks
Example 1 - The 6 lexical tones of Matbat -
Normalized time - Utterance-medial position,
following low target - Tracks averaged over 2
repetitions of 48 items uttered by 8 speakers.
(784 tokens)
55
F0 Analysis / Plotting
tracks
Example 2 - The 3 word-prosodic patterns of
Papiamentu. - Normalized time - Whole sentence
represented. - Tracks averaged over 2 repetitions
of 2 items uttered by 8 speakers. (96 tokens)
SUBJ COP O1 V1 O2 V2 PREP.
word-acc. I, penult. stress word-acc. II,
penult. stress word-acc. II, final stress
56
F0 Analysis / Plotting
tracks
Script pp_show_series10.psc (example) Function
on the basis of checked tracks, the scripts
produces a text file with an F0 values for each
of 8 points of measurement. Takes voicing at
edges into consideration.
57
Measuring overall / selective intensity (dB)
58
dB
Introduction
Variation in perceived voice quality (breathy,
modal, creaky) correlates with distribution of
energy in spectrum.
59
dB
Introduction
Functions include 1. Utterance-level
contrasts Example creaky voice correlates with
low F0 Q The slugs ate the dahlias, didnt
they? A No, thats not true / the rabbits ate
the dahlias, not the slugs. (Thanks to Mariko
Sugahara for the example)
60
dB
Introduction

• 2. Word-level contrasts on its own
• Dinka example breathy vs. modal
• ra??al ra??all l????tt l????t
• vein-sg. vein-pl. insult-sg.
  insult-pl.

61
dB
Introduction

• Functions include
• 2. Word-level contrasts on its own
• Dinka example breathy vs. modal
• ra??al ra??all l????tt l????t
• vein-sg. vein-pl. insult-sg.
  insult-pl.
• or as a package (register tone e.g. Mon-Khmer
  languages, Chamic languages)

62
dB
Introduction
Variation in perceived loudness correlates with
gt the distribution of energy in the spectrum
(spectral balance) gt overall intensity Functions
include gt Lexical stress (cf. Sluijter van
Heuven 1996) gt Phrasal accent (cf. Heldner 2003)
63
dB
Introduction
In summary gt Selective intensity marks
distinction in voice quality AND distinctions in
loudness. gt Loudness contrasts may also
correlate with overall intensity. gt It
remains unclear whether / to what extent
loudness and voice quality have separate
correlates.
64
dB Measuring overall
intensity
gt No need for checking. Automated procedure is
possible, cf. measurement of duration. gt
Important issue controlling for variation in
irrelevant factors in the course of session. gt
Relate intensity of target segment to the
intensity of (part of) the carrier utterance.
65
dB Measuring selective
intensity
Abundance of possible measurements, including gt
H1-H2 gt H1-A1, H1-A2, H1-A3 gt Dynamic filter
(Heldner) gt Average within a range (Sluijter
van Heuven)
See thematic issue of JPhon 294 (2001)
66
dB Measuring selective
intensity
Recommendations gt For detailed acoustic study
a measure of specific spectral properties is
best (explanatory adequacy). gt In relation to
a big corpus, Heldners filter- based measure
seems best (relatively vowel- independent easy
to automate) gt Try out several / Make your own
variation
67
dB Measuring selective
intensity
How to gt Point of measurement (cf. vowel
quality) gt Semi-automating measurements using
script
68
dB Measuring selective
intensity
Script msrcheck_spectr_indiv_interv.psc Goal
Semi-automated procedure for measurement of H1,
H2, A1, A2, A3. Extension of vowel quality
script.
69
gt The organizers, for inviting me.gt Thanks
to Patti Adank and Alice Turk, for discussions
on measurements of vowel quality and to Helen
Hanson, for discussions on voice quality. gt
The Netherlands Organization for Scientific
Research (NWO), for funding my research by means
of a postdoc grant to Vincent van Heuven.
Acknowledgements
70
Conference announcement Between Stress and
Tone Topic Typology of prosodic systems
When/where 16-18 of June 2005 / Leiden
Abstracts due 1 of November 2004
Details http//www.iias.nl/iias/agenda/best/