Conference PaperPDF Available

The Acoustics of Fricative Contrasts in Two German Dialects

Authors:

Abstract

In this study, we are investigating the acoustic characteristics of the five voiceless German fricatives [f s ç ʃ χ] elicited in non-words from 3 speakers each of two German dialects. The northern German dialect differentiates these five fricatives whereas in the middle German region and in Berlin, /ç/ and /ʃ/ have already merged or are in the process of merging [1;2]. Our previous work (submitted) has indicated that differentiating [ç] and [ʃ] acoustically in the speech of northern speakers from that of Berlin speakers works best when using DCT (discrete cosine transformation) rather than the four spectral moments. Results from our study corroborate this finding for both the northern and the middle German dialect.
The Acoustics of Fricative Contrasts in Two German Dialects
Stefanie Jannedy
1
, Melanie Weirich
2
1
ZAS Berlin, Germany
2
Friedrich-Schiller Universität Jena, Germany
jannedy@gmail.com, melanie.weirich@uni-jena.de
Abstract
In this study, we are investigating the acoustic characteristics
of the five voiceless German fricatives [f s ç ʃ χ] elicited in
non-words from 3 speakers each of two German dialects. The
northern German dialect differentiates these five fricatives
whereas in the middle German region and in Berlin, /ç/ and /ʃ/
have already merged or are in the process of merging [1;2].
Our previous work (submitted) has indicated that differentiat-
ing [ç] and [ʃ] acoustically in the speech of northern speakers
from that of Berlin speakers works best when using DCT (dis-
crete cosine transformation) rather than the four spectral mo-
ments. Results from our study corroborate this finding for both
the northern and the middle German dialect.
Index Terms: German fricatives, spectral moments, DCT
1. Introduction
Fricatives are produced by obstructing the airstream at some
place in the oral cavity, thereby creating turbulent noise. The
place of the constriction affects the spectral characteristics of
the noise: the more fronted a constriction is, the greater is the
energy in the higher frequencies. A constriction further back in
the oral cavity results in a spectrum with more energy in lower
frequency regions. Thus, energy peaks of the postalveolar fric-
ative /ʃ/ have lower frequencies than the energy peaks of the
alveolar fricative /s/ [3]. Moreover, the acoustic discrimination
of fricatives includes differences in intensity, spectral shape
(e.g. skewness, peakedness) and also temporal parameters.
Several studies have shown that fricative production is charac-
terized by a rather high inter-speaker variability especially in
sibilants cross-linguistically [4;5;6] mirroring both physiologi-
cal and social sources of variation in fine phonetic detail (e.g.
[7;8].
Numerous studies have investigated the various acoustic pa-
rameters potentially differentiating the fricatives of English
(e.g. [3;9;10;11;12;13;14;15]). Much fewer studies have exam-
ined fricative realization in other languages such as Aleut,
Scottish Gaelic or Chickasaw [6] or Polish [16;17;18]. Acous-
tic studies of the German fricative system are even more lim-
ited, [19] focused on the realization of German /h/ and /ç/ and
compared it with the Japanese /hi/-syllable and [20] (this vol-
ume) are exploring the German fricative system.
The German fricative system is rather interesting because this
system of contrasts involves the /ç/ sound which is relatively
rare in the languages of the world. The distribution of /ç/ is
restricted – it can only occur after high front vowels or word-
or morpheme initially. Moreover, German is one of only three
known languages of the world that contrasts the palatal frica-
tive /ç/ and the postalveolar fricative /ʃ/ [21].
This contrast however has already dissolved in the middle
German dialect region [22;23;24] and now the merger between
/ç/ and /ʃ/ is also affecting the speech of speakers in the north-
east of the middle German dialect belt up to Brandenburg and
Berlin [1;2].
Previous work [17;2] has shown that the four spectral mo-
ments 1. center of gravity (COG), 2. Standard Deviation from
the COG, 3. Skewness and 4. Kurtosis are less useful in differ-
entiating /ç/ and /ʃ/ in German, and that Discrete Cosine Trans-
formations (DCTs) [25] provide a useful method to differenti-
ate different fricative categories in other languages with com-
plex fricative systems such as Polish [16;17].
Thus, the purpose of the present study is to describe and eval-
uate the various acoustic parameters in terms of their useful-
ness to differentiate the contrasting five voiceless German
fricatives [f s ç ʃ χ]. Fricative productions will be compared
between speakers from the middle German dialect region
where the contrast between /ç/ and /ʃ/ has mostly dissolved in
spontaneous speech and speakers from the northwest, where
the contrast between /ç/ and /ʃ/ is still fully realized.
2. Methods
For this study, we have recorded and analyzed the data of 3
female speakers from the northern part of Germany and 3 fe-
males from the middle German dialect region around Jena. All
speakers were recorded with a head-mounted microphone at a
recording frequency of 44 kHz, however, data was
downsampled to 22kHz as fricative noise does not extend be-
yond 11 kHz. All speakers read the same list of 46 sentences
containing a carrier phrase and 46 different target words. The
reading of the list was repeated 4 times in the same order.
2.1. Recording Materials
The target words were either real German words or non-
words, made up to keep the segmental context identical. For
the purpose of this study, only the five non-words contrasting
in the fricatives are considered for analysis. The block of these
words were always read last.
1. Ich habe „iffa“ gesagt. /f/
2. Ich habe „issa“ gesagt. /s/
3. Ich habe „icha“ gesagt. /ç/
4. Ich habe „ischa“ gesagt. /ʃ/
5. Ich habe „acha“ gesagt. /χ/
2.1.1. Segmentation
A text file and a sound file for each repetition was submitted
to webMAUS [26] which generated a Praat TextGrid file
based on a first pass of a speech recognizer, trying to align the
word- and SAMPA canonical phonological transcriptions to
0 2000 4000 6000 8000 10000
-80
-70
-60
-50
-40
average fricative spectra JENA
c
sh
s
x
f
0 2000 4000 6000 8000 10000
-70
-60
-50
-40
-30
average fricative spectra BUX
c
sh
s
x
f
the audio file. The alignment of the boundaries was then hand
corrected for the target words and fricatives.
2.1.2. Measurements
All acoustic measurements were done in PRAAT [27]. For the
acoustic parameterization, the spectral moments (treating the
spectrum as a probability density distribution) following [11]
were calculated consisting of 1) the centroid frequency or Cen-
ter of Gravity (COG), , 2) the Standard Deviation (SD) which
is a measure of how much the frequencies in the spectrum de-
viate from the COG, 3) the skewness describing the energy
distribution over the whole frequency range of the spectrum
and expresses if the frequencies are skewed towards the higher
or the lower frequencies; and 4) kurtosis which reveals the
spectral peakedness of the distribution. In addition, Discrete
Cosine Transformation (DCT) [28] was used to quantify the
shape of the spectra. DCT decomposes the signal into a set of
half-cycle cosine waves whereby the resulting amplitudes of
these cosine waves are the DCT coefficients. We will concen-
trate on three DCT coefficients, which 1) are proportional to
the linear slope of the spectrum (DCT1), 2) correspond to its
curvature (DCT2), and 3) describe the amplitude of the higher
frequencies (DCT3). Prior to analysis the data was filtered us-
ing a pass band filter (200-11025 Hz) and all acoustic meas-
urements were automatically logged at the temporal midpoint
of the fricatives.
For a better quantification of the fricative contrasts, Euclidean
Distances (EDs) were calculated between all fricative pairs for
each speaker separately using 1) the spectral moments and 2)
the DCT1xDCT2xDCT3 space.
2.2. Speakers
For this analysis we used the data from 3 female speakers for
each dialect group. The northern dialect speakers were born,
raised and now live just to the south of Hamburg. They range
in age between 39 and 44. The middle German speakers were
born and raised in Thuringia, now living in Jena. They range
in age between 23 and 34.
2.3. Statistics
All in all, 6 speakers x 4 repetitions x 5 fricatives = 120 items
were analyzed. We performed a linear mixed effects analysis
as implemented in the lme4 package [29] in R (version 2.14.1,
R Development Core Team 2008). Likelihood ratio tests were
run to test for a significant effect of the test variables by com-
paring the model with the factor in question to a model with-
out that factor. Post hoc tests were carried out (using the
lsmeans package in R [30]) to reveal any significant difference
between the dialect areas for each fricative or fricative pair
respectively.
3. Results
First, we will have a look on the average spectra of the five
fricatives for both dialect areas. Then, the acoustic space of the
fricatives is shown characterized by the different parameters
(COG, SD, skewness, kurtosis and DCT1-3). The statistical
analysis concentrates on investigating the different parameters
in their usability to distinguish the German fricatives and to
compare the realization of the fricative contrasts in the two
dialect areas.
3.1. Fricative space
Figure 1 shows the average fricative spectra for the five frica-
tives for the speakers from Bux (top panel) and from Jena
(lower panel). In the two dialect areas the five spectra vary in
the energy distribution over the frequency range in a similar
way. As expected /s/ (green) is characterized by high energy in
the higher frequency regions over 6000 Hz, while /ç/ (red) and
/ʃ/ (blue) have their energy peaks somewhat lower (for Bux
around 3000 Hz, for Jena around 5000 Hz). For Jena, no clear
peak but rather a plateau can be seen. Both /χ/ (magenta) and
/f/ (orange) show flatter spectra, but this is especially the case
for /f/ for the speakers from Bux.
Figure 1: Mean fricative spectra separated by dialect area,
different fricatives are plotted in different colors
Different measurements were made to see which of them are
best qualified to describe and separate the spectral characteris-
tics of the five German fricatives and which are maybe less
suitable. The upper plot of Figure 2 shows the acoustic space
of the fricatives spanned by the two spectral moments COG
and skewness. While /s/ and /χ/ are nicely separated by COG
(plotted on the x-axis), with /s/ (green) having high values and
/χ/ (magenta) low values, the values for the three fricatives /ç ʃ
f/ overlap. The same holds for skewness, which separates /s/
and /χ/ but fails to show dividable distributions for /ç/, /ʃ/ and
/f/. The two dialect areas show similar distributions, however,
overall, the values for skewness and COG are slightly shifted
with higher values for COG and lower values for skewness for
Jena than for Buxtehude.
The lower plot of Figure 2 shows the acoustic space of the
fricatives spanned by DCT1 and DCT2: a clearer separation
between the five fricatives can be seen for both dialect areas.
However, /ç/ remains in the middle of the fricative cloud,
thereby revealing the least clear separation to all other frica-
tives. Again, overall, the fricative distributions are somewhat
shifted between the dialect areas.
χ
χ
χ
χç
çç
fff
fʃ
ʃ
ʃ
ʃ
s
s
ss
χ
χ
χχç
ç
ç
f
ffʃ
ʃ
ʃʃs
ss
s
χ
χ
χ
χ
ç
ç
çf
fff
ʃʃʃ
ʃs
s
ss
χχχ
χçç
ç
çff
f
fʃ
ʃʃ
ʃs
s
s
s
χχ
χχ ç
ç
ç
çf
fff
ʃ
ʃʃ
ʃ
ss
s
s
χ
χχ
χççç
ç
ff
ff
ʃ
ʃʃ
ʃss
s
s
Buxtehude Jena
-2
0
2
4
2000 4000 6000 8000 2000 4000 6000 8000
CoG
skewness
χχ
χ
χ
çç
ç
f
f
ff
ʃʃ
ʃʃ
ss
s
s
χχ
χ
χ
ç
ç
ç
ff
f
ʃʃʃʃ
ss
s
s
χ
χ
χχ
çç
ç
ff
ff
ʃ
ʃ
ʃ ʃ
ss
s
sχ
χ
χχ
ç
ççç
f
ff
f
ʃ
ʃʃ
ʃ
ss
s
sχχχ
χ
çç
ç
ç
ff
f
f
ʃ
ʃ
ʃʃ
s
s
s
sχ
χχ
χ
ç
çç
ç
f
ff
f
ʃ
ʃ
ʃʃ
s
ss
s
Buxtehude Jena
-15
-10
-5
0
-10 -5 0 5 10 15 -10 -5 0 5 10 15
DCT1
DCT2
Figure 2: Acoustic fricative space separated by dialect area
using COG and skewness (upper plot) and DCT1 and DCT2
(lower plot).
Different statistical models were run for the different parame-
ters as dependent variable. As fixed effects, repetition was en-
tered as a control variable, and dialect area (Buxtehude, Jena)
and fricative (f, s, ç, ʃ, χ) as test variables. As random effect
we entered an intercept for speaker.
Table 1: Results of the Post Hoc Test showing significant dif-
ferences between the fricatives in DCT1, DCT2 and DCT3 for
the two dialect areas (*** p < 0.001, ** p < 0.01, * p < 0.05).
DCT1 DCT2 DCT3
pair
Bux
Jena
Bux
Jena Bux
Jena
/ç-f/
n.s. n.s. ***
*** ***
n.s.
/ç-s/
***
***
n.s. *** ***
***
/ç-ʃ/
* n.s. ***
*** n.s. n.s.
/ç-χ/
***
***
***
*** n.s. n.s.
/f-s/
***
***
***
n.s. ***
***
/f-ʃ/
***
n.s. ***
*** ***
*
/f-χ/
***
***
n.s. n.s. ***
n.s.
/s-ʃ/
***
***
***
*** ***
***
/s-χ/
***
***
***
n.s. ***
***
/ʃ-χ/
** ***
***
*** n.s. n.s.
For all DCT coefficients we found a significant interaction of
dialect area and fricative (p < 0.001). Results of Post Hoc
Tests analyzing the differences between the fricatives for the
two dialect areas are given in Table 1 (p values are adjusted
using the tukey method). Overall, many of the comparisons
were significant. While DCT3 shows fewer significant differ-
ences compared to DCT1 and DCT2, DCT3 clearly distin-
guishes /s/ from all other fricatives. Since DCT3 reflects the
energy of the higher frequencies, it works best capturing the
spectral characteristics of /s/. The fricative pairs most difficult
to differentiate are /ç-f/, /ç-ʃ/ and /f- χ/, while the sibilants /s/
and /ʃ/ are distinguished best by the DCT parameters.
When comparing the two cities, it can be seen that Buxtehude
shows more significant differences than Jena, thereby reveal-
ing a more distinct acoustic realization of the fricative con-
trasts.
For SD, skewness and kurtosis there was also a significant in-
teraction of dialect area x fricative, for COG however, there
was a main effect of fricative and of dialect area but no inter-
action thereof. The main effect of dialect area reflects a higher
COG for Jena than for Buxtehude already apparent in the
shifted distribution in Figure 2. In comparison to the DCT
analysis, fewer fricative comparisons turned out to be signifi-
cant. Moreover, for the /ç-ʃ/-contrast none of the parameters
succeeded in distinguishing the spectral characteristics of the
two fricatives. Parallel to the DCT analysis, more significant
differences were found for Buxtehude than for Jena.
Table 2: Results of the Post Hoc Test showing significant dif-
ferences between the fricatives in COG, SD, skewness and kur-
tosis for the two dialect areas (*** p < 0.001, ** p < 0.01, * p
< 0.05).
COG SD skewness
kurtosis
pair Bux Jena Bux Jena
Bux Jena Bux
Jena
/ç-f/ n.s. n.s. *** ***
n.s. n.s. n.s. n.s.
/ç-s/ *** *** *** n.s. *** n.s. n.s. n.s.
/ç-ʃ/ n.s. n.s. n.s. n.s. n.s. n.s. n.s. n.s.
/ç-χ/ *** *** n.s. ** *** *** * n.s.
/f-s/ *** *** *** ***
*** n.s. n.s. **
/f- ʃ /
n.s. n.s. *** ***
n.s. n.s. * n.s.
/f-χ/ *** *** *** ***
*** *** *** n.s.
/s-ʃ / *** *** *** n.s. *** *** n.s. **
/s-χ/ *** *** *** ***
*** *** *** *
/ʃ -χ/ *** *** n.s. n.s. *** n.s. ** n.s..
3.2. ED of Fricative contrasts.
Since the above analysis showed that the DCT coefficients are
better descriptors distinguishing the acoustic characteristics of
the five German fricatives we will concentrate on these pa-
rameters. Figure 3 shows the amount of the acoustic contrast
between all fricative pairs expressed as the Euclidean Distance
using DCT1-3 (ED_DCT123). The different dialect areas are
plotted next to each other. A significant interaction of pair x
dialect area was found (χ²(9) = 26.213, p = 0.0019). However,
it only reveals that the contrasts between /s-ʃ/ (p < .05) and /s-
χ/ (p < .001) were significantly higher in Jena than in Buxte-
hude. Overall, the order of the fricative pairs is much the same
in both dialect areas: The smallest acoustic contrast is revealed
for the /ç-ʃ/ pair; followed by the other /ç/-comparisons except
the one with /s/.
4. Discussion
While generally, DCTs were better than the four spectral mo-
ments in separating the fricatives in each dialect, especially the
curvature of the spectrum (DCT2) is a rather promising acous-
tic correlate for fricative differentiation, particularly when the
very similar spectral shapes of /ç/ and /ʃ/ are concerned. Alt-
hough the experiment was set up for speakers to visually cap-
ture the differences between the experimental tokens, and pre-
sumably all speakers made an effort to realize the acoustic
contrasts reflected in the orthography, fewer fricative pairs
turned out to be significantly different for the Jena speakers
compared to the northern speakers. This is not surprising given
the overall tendency of Thuringian speakers to merge at least
/ç/ and /ʃ/ in fluent and spontaneous speech.
In light of this, it is interesting that the rare phoneme category
/ç/ even evolved as a contrastive entity in a crowded fricative
space as found in German. While our results capture the ob-
servation that the /ç-ʃ/ contrast is difficult to quantify and qual-
ify even in the northern speakers, it is not a surprise that it is
on the demise in the middle German Thuringian dialect. A lis-
tening experiment is planned to analyze if and how the con-
trast is perceived in the different areas. It remains to be seen if
the /ç-ʃ/ merger also starts spreading further north. However,
there may be some resistance by northern speakers which is
not to sound like “southern” speakers.
Figure 3: Acoustic contrast between fricatives estimated as EDs in DCT1 x DCT2 x DCT3 space separated by dialect area.
5. Acknowledgements
This work was funded by the German Bundesministerium für
Bildung und Forschung (BMBF) (Grant Nr. 01UG1411). We
would also like to thank Adrian Simpson (F.S. Univ. Jena) for
allowing us to use his praat scripts for parameter extractions,
Sophie Arndt for her patient work as a research assistant, and
all of our informants.
6. References
[1] Jannedy, S. & Weirich, M. (2014). Perceptual divergence in an
urban setting: category instability of the palatal fricative. Journal
of Laboratory Phonology 5(1):91-122.
[2] Jannedy, S. & Weirich, M. (submitted) Spectral Differentiation
of Intradialectal Fricative Variation. Submitted to Clopper, C.
(ed). Special Issue of JASA on Methods in Dialect Research.
[3] Hughes, G. & Halle, M. (1956). Spectral properties of fricative
consonants. Journal of the Acoustical Society of America 28, pp.
303-310.
[4] Dart, S. N. (1998) Comparing French and English coronal con-
sonant articulation. Journal of Phonetics 26, 71-94.
[5] Newman, R. S.; Clouse, S. A. & Burnham, J. L. (2001) The per-
ceptual consequences of within-talker variability in fricative
production. Journal of the Acoustical Society of America 109(3),
1181-1196.
[6] Gordon, M.; Barthmaier, P. & Sands, K. (2002) A cross-
linguistic acoustic study of voiceless fricatives. Journal of the In-
ternational Phonetic Association 32, 141-174.
[7] Weirich, M. & Fuchs, S. (2013) Palatal morphology can influ-
ence speaker-specific realizations of phonemic contrasts. Journal
of Speech, Lanuage and Hearing Research 56, S1894–S1908.
[8] Stuart-Smith, J. (2007): Empirical evidence for gender speech
production: /s/ in Glaswegian. In: Cole, J. & Hualde, J. I. (eds.):
Laboratory Phonology 9. Berlin: Mouton. pp. 65-86.
[9] Harris, K. S. (1958): Cues for the discrimination of American
English fricatices in spoken syllables. Language and Speech 1.
pp. 1-7.
[10] Shadle, C. (1985) The acoustics of fricative consonants. Tech-
nical Report 506, Research Laboratory of Electronics, MIT
Cambridge.
[11] Forrest, K., Weismer, G., Milenkovic, P. & Dougall, R.N.
(1988). Statistical analysis of word-initial voiceless obstruents:
Preliminary data. Journal of the Acoustical Society of America
841, 115-123.
[12] Behrens, S. J. & Blumstein, S. E. (1988a). Acoustic characteris-
tics of English voiceless fricatives: A descriptive analysis. Jour-
nal of Phonetics 18, 51-63.
[13] Behrens, S. J. & Blumstein, S. E. (1988b). On the role of the
amplitude of the fricative noise in the perception of place of ar-
ticulation in voiceless fricative consonants. Journal of the Acous-
tical Society of America 84, 861-867.
[14] Tomiak, G. R. (1990). An acoustic and perceptual analysis of the
spectral moments invariant with voiceless fricative obstruents.
Doctoral dissertation, SUNY Buffalo.
[15] Jongman, A.; Wayland, R. & Wong, S. (2000). Acoustic charac-
teristics of English fricatives. Journal of the Acoustical Society
of America 108, 1252-1263.
[16] Bukmaier, V. & Harrington, J. (2016). The articulatory and
acoustic characteristics of Polish sibilants and their consequenc-
es for diachronic change. Journal of the International Phonetic
Association, pp. 1-19. doi:10.1017/S0025100316000062.
[17] Guzik, K. & Harrington, J. (2007). The quantification of place of
articulation assimilation in electropalatographic data using the
similarity index (SI). Advances in Speech Language Pathology 9
(1), 109-119.
[18] Wiktor, J. (1995). The acoustic parameters of Polish voiceless
fricatives: Analysis of variance. Phonetica 52:252-258.
[19] Tronnier, M. & M. Dantsuji. 1993. An Acoustic Approach to
Fricatives in Japanese and German. Proceedings of the 3rd
EUROSPEECH'93, Berlin, vol. 1, 271-274.
[20] Lowery, M. & Kleber, F. An acoustic analysis of German frica-
tives. (this volume)
[21] Mielke, J. (2008). The emergence of distinctive features, pp. 1-
304. Oxford: Oxford University Press.
[22] Herrgen, J. (1986). Koronalisierung und Hyperkorrektion. Das
palatale Allophon des /CH/-Phonems und seine Variation im
Westmitteldeutschen, pp. 1-278. Stuttgart: Franz Steiner.
[23] Dirim, I. & Auer, P. (2004). Türkisch sprechen nicht nur die
Türken. Über die Unschärfebeziehung zwischen Sprache und
Ethnie in Deutchland. Walter De Gruyter, pp 1-255.
[24] Hall, T. A. (2013). Alveolopalatalization in Central German as
markedness reduction. Transactions of the Philological Society,
pp. 143-166. doi: 10.1111/1467-968X.12002.
[25] Harrington, J. (2010). Phonetic analysis of speech corpora, pp.
1-424. Chichester: Wiley-Blackwell.
[26] Kisler, T. and Reichel U. D. and Schiel, F. and Draxler, Ch. and
Jackl, B. and Pörner, N. (2016): BAS Speech Science Web Ser-
vices - an Update of Current Developments, Proceedings of the
10th International Conference on Language Resources and Eval-
uation (LREC 2016), Portorož, Slovenia, paper id 668.
[27] Boersma, P. and D. Weenink (2012). Praat: doing phonetics by
computer [Computer program]. Version 5.3.23, retrieved 7 Au-
gust 2012 from http://www.praat.org/
[28] Watson, C. I. & Harrington, J. (1999). Acoustic evidence for
dynamic formant trajectories in Australian English vowels. Jour-
nal of the Acoustical Society of America 106, 458-468.
[29] Bates, D., Maechler, M. & B. Bolker (2011). lme4: Linear
mixed-effects models using S4 classes, R package version
0.999375-42.
[30] Lenth, R. V. (2016) Least-Squares Means: The R Package
lsmeans. Journal of Statistical Software, 69(1), 1-33.
doi:10.18637/jss.v069.i01
Buxtehude Jena
5
10
15
20
25
çʃçf çχfχfs fʃ χʃ sç sʃsχçʃçf çχfχfs fʃ χʃ sç sʃsχ
fricative pair
ED_DCT123
... Weirich (2012) in her dissertation showed that DCT were useful in describing speakerspecific implementations of /s/ and /S/ in the speech of monozygotic and dizygotic twins. We have previously done some pilot work on the acoustics of /ç/ and /S/ in northern Standard German, eastern Middle German (Jannedy and Weirich, 2016), and in Berlin German (Jannedy et al., 2015) that points to DCT as a parameter that is able to differentiate the very similar spectra of these fricatives. More detail on the role of DCT in the characterization and comparison of fricative spectra is given in Sec. ...
... DCT, a method proposed by Watson and Harrington (1999), was used to quantify the shape of the spectra and, in particular, the fricative contrast in more detail (using MATLAB). For a better understanding of the spectral parameterization using DCT, we will have a look at the fricatives /ç/ and /S/ produced by speakers from Buxtehude (Jannedy and Weirich, 2016), located in the northern dialect region where the contrast is fully realized. This contrast is mirrored by the difference in spectral shape between the palatal [ç] (gray) and postalveolar [S] (black) in Fig. 3. ...
... FIG. 3. Averaged fricative spectra of /ç/ (gray) and /S/ (black) of speakers from Buxtehude with cosine waves corresponding to DCT1 (left) and DCT2 (right). (Jannedy and Weirich, 2016). DCT1 reflects the slope of the spectrum and thus results in higher values for [S] (black) than for [ç] (gray), since the difference between the energy distribution of the lower frequencies (left of the midpoint of the spectrum marked by the crossing lines) and the higher frequencies (right of the midpoint) is higher for [S] than for [ç]. ...
Article
Full-text available
The results of the quantification of the acoustic differences between German /ç/ and /ʃ/ in three speaker groups with varying contrast realizations are presented. Data for two speaker groups were collected in Berlin and Kiel, where the contrast is still realized. Data for a third group were collected in Berlin from speakers of Hood German—a youth-style multiethnolect spoken by adolescents in multilingual and multicultural neighborhoods of Berlin—where the contrast has weakened or is even lost. A forced choice perception test showed that listeners reliably differentiate these two fricatives in minimal pairs produced by the speakers from Berlin and Kiel, but fail to do so for the productions of the Hood German speakers. The acoustic analysis reveals that spectrally, the fricatives are very similar in all varieties. The spectral moments (Center of Gravity, standard deviation, kurtosis, skewness) fail to reveal the differences between the fricatives that are apparent from visual inspection of the spectra and the perceived auditory differences. Analyses of the discrete cosine transformation coefficients, however, better quantify these differences. This study suggests that minute differences between fricatives that vary between speaker groups may be captured more reliably with discrete cosine transformations compared to spectral moments.
... DCT1 is proportional to the linear slope of the spectrum, DCT2 corresponds to its curvature and DCT3 describes the amplitude of the higher frequencies. DCTs have been shown to provide a very effective separation between the four fricative types in Polish [18], and to be a reliable parameter to differentiate the very similar acoustic spectra of /ç/ and /ʃ/ in two dialectal regions in Germany [19]. ...
Conference Paper
Full-text available
Differences between male and female speakers have been explained in terms of biological inevitabilities but also in terms of behavioral and socially motivated factors. The aim of this study is to investigate the latter by examining gender-specific variability within the same gender. The speech of 29 German men and women-all of them expecting their first child but varying in the time they plan to stay at home during their child's first year (parental role)-is analyzed. Acoustic analyses comprise the vowel space size and the realization of the inter-sibilant contrast. While the data is part of a larger longitudinal project investigating adult-and infant-directed speech during the infant's first year of life, this study concentrates on the recordings made before the birth of the child. Inter-speaker variability is investigated in relation to 1) the chosen parental role and 2) self-ascribed ratings on positive feminine attributes (gender identity). Results show that both factors (planned duration of parental leave and the femininity ratings) contribute to the variability found between, but also within the same gender. In particular, the vowel space size was found to be positively correlated with self-ascribed femininity ratings in male speakers.
... The spectral moments consist of 1) the centroid frequency or Center of Gravity (COG), 2) the Standard Deviation (SD) of the COG, 3) the skewness describing the energy distribution over the whole frequency range of the spectrum and expresses if the frequencies are skewed towards the higher or the lower frequencies; and 4) kurtosis which reveals the spectral peakedness of the distribution. Recently, DCT has been shown to distinguish sibilants very well ( [21,22,23,24]). DCT decomposes the signal into a set of half-cycle cosine waves whereby the resulting amplitudes of these cosine waves are the DCT coefficients. We will concentrate on three DCT coefficients, which 1) are proportional to the linear slope of the spectrum (DCT1), 2) correspond to its curvature (DCT2), and 3) describe the amplitude of the higher frequencies (DCT3). ...
Conference Paper
Full-text available
We are investigating the speech of German and Swedish mothers and fathers during the first year of their first baby. Both infant-and adult-directed speech are analyzed and compared between the sexes but also between different time points during the first year. In addition, the involvement in child care is considered as a potential factor. We are now in the process of finding participants and gathering data (read speech and spontaneous speech) from the first recording before birth of the child. Here, the speech material, our hypotheses and first pilot studies are presented.
Article
Full-text available
The present electroencephalographical multi-speaker MMN oddball experiment was designed to study the phonological processing of German native and non-native speech sounds. Precisely, we focused on the perception of German /ɪ-iː/, /ɛ-ɛː/, /a-aː/ and the fricatives [ʃ] and [ç] in German natives (GG) and French learners of German (FG). As expected, our results showed that GG were able to discriminate all the critical vowel contrasts. In contrast, FG, despite their high L2 proficiency level, were only marginally sensitive to vowel length variations. Finally, neither GG nor FG discriminated the opposition between [ʃ] and [ç], as revealed by the absence of MMN response. This latter finding was interpreted in terms of low perceptual salience. Taken together, the present findings lend partial support to the Perceptual Assimilation Model for late bilinguals (PAM-L2) for speech perception of non-native phonological contrasts.
Conference Paper
Full-text available
In 2012 the Bavarian Archive for Speech Signals started providing some of its tools from the field of spoken language in the form of Software as a Service (SaaS). This means users access the processing functionality over a web browser and therefore do not have to install complex software packages on a local computer. Amongst others, these tools include segmentation & labeling, grapheme-to-pho-neme conversion, text alignment, syllabification and metadata generation, where all but the last are available for a variety of languages. Since its creation the number of available services and the web interface have changed considerably. We give an overview and a detailed description of the system architecture, the available web services and their functionality. Furthermore, we show how the number of files processed over the system developed in the last four years.
Article
Full-text available
Least-squares means are predictions from a linear model, or averages thereof. They are useful in the analysis of experimental data for summarizing the effects of factors, and for testing linear contrasts among predictions. The lsmeans package (Lenth 2016) provides a simple way of obtaining least-squares means and contrasts thereof. It supports many models fitted by R (R Core Team 2015) core packages (as well as a few key contributed ones) that fit linear or mixed models, and provides a simple way of extending it to cover more model classes.
Article
Full-text available
Purpose: The purpose of this study was to further explore the understanding of speaker-specific realizations of the /s/-/ʃ/ contrast in German in relation to individual differences in palate shape. Method: Two articulatory experiments were carried out with German native speakers. In the first experiment, 4 monozygotic and 2 dizygotic twin pairs were recorded by means of electromagnetic articulography. In the second experiment, 12 unrelated speakers were recorded by means of electropalatography. Interspeaker variability in the articulatory distance between the sibilants was measured and was correlated with several parameters of the palate shape. Results: The results were twofold: (a) Similar palatal morphologies as found in monozygotic twins yield similar articulatory realizations of the /s/-/ʃ/ contrast regarding vertical and horizontal distance of the target tongue tip positions, and (b) the realization of the contrast was influenced by palatal steepness, especially the inclination angle of the alveolo-palatal region. Speakers with flat inclination angles mainly retracted their tongue to realize the contrast, whereas speakers with steep inclination angles also elevated their tongue. Conclusion: The articulatory realization of the sibilant contrast is influenced not only by speaker-specific auditory acuity, as previously observed, but also by palatal shape morphology, which affects the somatosensory feedback speakers receive.
Article
Full-text available
The differential categorization of identical stimuli depending on the presence of a prime is described as a perceptual divergence effect. We examined whether native listeners of the Berlin vernacular of German categorized identical acoustic stimuli differently in the explicit context of the names of two different districts of Berlin, assuming that listeners infer social information and linguistic stereotypes based on the names of these neighborhoods (Kreuzberg vs. Zehlendorf). All listeners categorized natural acoustic stimuli with synthetic fricatives synthesized along a continuum ranging from /ç/ to /ʃ/ as either Fichte /fɪçtə/ (`spruce') or fischte /fɪʃtə/ (1st person sg. `to fish'). This variable was chosen because many young multiethnic speakers of Berlin German pronounce /ç/ as [ʃ] or [ɕ], and this alternation is highly associated with speakers with a migrant background from Kreuzberg. Data were gathered in a forced-choice identification task, and, for a subset of the participants, reaction times (RTs) were also gathered. Results indicate a differential categorization pattern depending on (1) the copresented information, i.e., Kreuzberg, Zehlendorf, or none (control), and (2) the age of the listeners, with older listeners being more affected by the co-presented information. While older listeners categorized significantly more /ʃ/ sounds in the context of Kreuzberg than in the control or Zehlendorf condition, younger listeners rated most /ʃ/ sounds in the control condition (no added information). Results are interpreted in terms of a potential sound change in progress: the loss of the phoneme contrast between /ç/ and /ʃ/.
Article
The study is concerned with the relative synchronic stability of three contrastive sibilant fricatives / s ʂ ɕ/ in Polish. Tongue movement data were collected from nine first-language Polish speakers producing symmetrical real and non-word CVCV sequences in three vowel contexts. A Gaussian model was used to classify the sibilants from spectral information in the noise and from formant frequencies at vowel onset. The physiological analysis showed an almost complete separation between / s ʂ ɕ/ on tongue-tip parameters. The acoustic analysis showed that the greater energy at higher frequencies distinguished / s / in the fricative noise from the other two sibilant categories. The most salient information at vowel onset was for /ɕ/, which also had a strong palatalizing effect on the following vowel. Whereas either the noise or vowel onset was largely sufficient for the identification of / s ɕ/ respectively, both sets of cues were necessary to separate /ʂ/ from / s ɕ/. The greater synchronic instability of /ʂ/ may derive from its high articulatory complexity coupled with its comparatively low acoustic salience. The data also suggest that the relatively late stage of /ʂ/ acquisition by children may come about because of the weak acoustic information in the vowel for its distinction from / s /.
Article
by Christine Helen Shadle.
Article
Averaged spectra of the Polish voiceless fricatives are described in terms of three central moments, and the variables are analysed using both univariate and multivariate analysis of variance. The vowel context effect is strong, but that of the (male) speaker is not. /∫/ is not distinct from /ҩ/ in this experiment, but this failure should not be a handicap in automatic speech recognition.