Eppur si muove: Formant Dynamics is Relevant for The Study of Speech

Aging Effects

Luciana Albuquerque

1,2,3,4 a

, Catarina Oliveira

1,5 b

, Ant

onio Teixeira

1,3 c

and Daniela Figueiredo

2,5 d

Institute of Electronics and Informatics Engineering of Aveiro, University of Aveiro, Aveiro, Portugal

Center for Health Technology and Services Research, University of Aveiro, Aveiro, Portugal

Department of Electronics Telecommunications and Informatics, University of Aveiro, Aveiro, Portugal

Department of Education and Psychology, University of Aveiro, Aveiro, Portugal

School of Health Science, University of Aveiro, Aveiro, Portugal

Keywords:

Aging Speech, Vowels, Dynamic Formant Frequencies, European Portuguese.

Abstract:

The evidence have shown that speech change with age and the automatic speech recognition systems needs

adaptation to older voices. Most of the acoustic studies about the age effects on speech production have

focused on static approaches to obtain the vowel formants. However, vowel formant dynamics may also be

important to characterize vowel quality and the age related changes. In this position paper the authors argue

for the need to increase the use of dynamic information in acoustic studies. Among the main arguments, we

can state that: speech is inherently dynamic; dynamic vowel formants improve the classiﬁcation of vowels and

dialects and play an important role in vowel perception; nowadays better tools allow to go beyond analysis of

snapshots.

1 INTRODUCTION

As speech is a physiological signal that provides in-

formation at multiple levels concerning the linguistic

aspects (e.g. words, message, accent, language) as

well as the paralinguistic characteristics (e.g. gender,

age, emotional state) (Sadjadi et al., 2016; Yue et al.,

2014; Qawaqneh et al., 2017), the human speech can

be used as an important cue to represent the person’s

age (Yue et al., 2014; Sch

otz, 2006).

The age effects on speech production mechanism

have a signiﬁcant impact on the acoustic measure-

ments of speakers’ vocal output (Xue and Hao, 2003;

Sch

otz, 2006; Braun and Friebis, 2009). Despite

the wide range of speech acoustic measurements that

could be affected by age, in this position paper we will

focus on vowel acoustics, namely on the different ap-

proaches to study the vowel formant frequencies.

https://orcid.org/0000-0003-1654-3272

https://orcid.org/0000-0002-3389-3082

https://orcid.org/0000-0002-7675-1236

https://orcid.org/0000-0002-3160-7871

1.1 Background Information

Population aging, while due primarily to lower fertil-

ity, also reﬂects a human success story of increased

longevity (He et al., 2016). According to the World

Health Organization (2012a,b) the number of peo-

ple aged over 65 is increasing and Portugal is one

of the developed countries with the highest rate of

older population. Between 1970 and 2018, the per-

centage of people aged 65 and over increased from

9.7% to 21.8% Statistics Portugal (2019, 2015). This

age group may increase from 2.1 million to 2.8 mil-

lion between 2015 and 2080 in Portugal.

However, increasing longevity lead to new chal-

lenges, such as a pressure on health care costs, achiev-

ing life expectancy in good health, living indepen-

dently Makiyama and Hirano (2017); He et al. (2016).

Aging involves changes at physiological, cogni-

tive, psychological and social levels. Physiological

age-related changes take place in different tissues and

organs, and the human speech production mechanism

is no exception Makiyama and Hirano (2017); Braun

and Friebis (2009). Not only do the cognitive skills

which are required in the planning process change

with age, but also do initiation, phonation and artic-

276

Albuquerque, L., Oliveira, C., Teixeira, A. and Figueiredo, D.

Eppur si muove: Formant Dynamics is Relevant for The Study of Speech Aging Effects.

DOI: 10.5220/0010320902760283

In Proceedings of the 14th International Joint Conference on Biomedical Engineering Systems and Technologies (BIOSTEC 2021) - Volume 4: BIOSIGNALS, pages 276-283

ISBN: 978-989-758-490-9

ulation Braun and Friebis (2009).

Automatic extraction of speaker dependent char-

acteristics, from a short speech utterance, is a chal-

lenging task, and it has a wide range of applica-

tions (Barkana and Zhou, 2015; Sadjadi et al., 2016;

Qawaqneh et al., 2017). The age and gender detection

of a speaker is a rapidly emerging ﬁeld of research due

to the continually growing interest in applications of

communication, human–computer interface and natu-

ral spoken-dialog systems Barkana and Zhou (2015).

Overall, due to the aging speech alterations, the

current automatic speech recognition (ASR) systems

still do not work well with older people’s speech

ainen et al. (2012); Pellegrini et al. (2013). Be-

sides, older adults have more difﬁculties in the in-

teraction with computers, which represents a barrier

to their access to new technologies H

ainen et al.

(2012).

Although previous studies have presented some

progress in extract and select features that repre-

sent the speaker’s characteristics uniquely, the clas-

siﬁcation of speaker’s age still has a lot to improve

Qawaqneh et al. (2017), which is essential for the de-

velopment of ASR systems suitable for older voices

Vipperla et al. (2010). Moreover, a deeper knowledge

of how speech changes with age is essential for clini-

cal assessment and treatment of different speech dis-

orders (Sataloff et al., 1997; Johns III et al., 2011),

and also to provide information for other ﬁelds (e.g.,

phonetics, speech science, forensic linguistics and

biometric recognition) (Kent and Vorperian, 2018;

Lanitis, 2010).

2 STATE OF THE ART: AGING

ACOUSTIC CHANGES

During the natural process of aging anatomical and

physiological changes in the speech organs occur,

which are reﬂected in the variation of several acous-

tic parameters, more speciﬁcally in the decrease of

the speaking rate, in the increase of speech pauses,

in the variation of the fundamental frequency and in

the changes in the pattern of the formant frequen-

cies, among others Linville and Rens (2001); Sch

otz

(2006).

The magnitude of the speech acoustic changes

depends upon the individual, as the voice is intri-

cately linked to the dynamics of the speech organs

Makiyama and Hirano (2017). Furthermore there are

substantial gender differences in the extent and timing

of the aging process Linville (2001); Makiyama and

Hirano (2017); Sch

otz (2006).

Numerous studies have focused on age-related

changes on F1 and F2, neglecting higher formants,

and using a static approach to obtain the formant

frequencies.Moreover, the results across studies are

highly inconsistent. There are studies that have

shown an age-dependent formant frequency lower-

ing Linville and Rens (2001); Xue and Hao (2003);

Watson and Munson (2007); Harrington et al. (2007);

Decoster and Debruyne (1999) and others that have

reported no changes in formant frequencies Fletcher

et al. (2015); Sebastian et al. (2012). In Sch

otz (2006)

and Eichhorn et al. (2018) there are vowels that pre-

sented a different pattern of formant frequencies vari-

ation with age and gender. In addition, some studies

have referred a centralization of the vowel space in

older speakers, which should result in a movement to

the centroid of formant space Sch

otz (2006); Rastat-

ter and Jacques (1990); Rastatter et al. (1997); Torre

III and Barlow (2009); Mertens et al. (2020).

Concerning temporal information, it has often

been noted that older adults use slower speaking rates

(Linville, 2001; Sch

otz, 2006) and therefore produce

vowels with longer durations Linville (2001); Sch

otz

(2006); Fletcher et al. (2015); Benjamin (1982);

Fougeron et al. (2018).

To the best of our knowledge, there are a few

data available about the age effects on dynamic for-

mant frequencies. Only Jacewicz et al. (2011a), in a

cross-generational and cross-dialectal study, revealed

substantial differences both in formant dynamics and

vowel dispersion in the acoustic space as a function

of age group.

2.1 European Portuguese Vowels

Most previous studies on European Portuguese (EP)

vowels rely on single acoustic measurements of for-

mant frequencies at the middle of the vowel (Pelle-

grini et al., 2013; Albuquerque et al., 2014, 2019;

Oliveira et al., 2012).

The few data available for the EP have shown

that age changes in vowel formant frequencies are

not consistent and seem to be different among vowels

(Albuquerque et al., 2019, 2014). These age-related

acoustic changes might occur due to speciﬁc articula-

tory adjustments of the older speakers during speech,

rather than generalized processes such as lengthen-

ing of the vocal tract (Xue and Hao, 2003; Eichhorn

et al., 2018). Additionally, Albuquerque et al. (2019)

showed a trend towards the centralization of vowels’

space with age, mainly for males. This might indicate

that articulation capability of males deteriorate with

aging (Arias-Vergara et al., 2017). Further, vowel du-

ration has presented consistent results in EP, show-

ing a signiﬁcantly increase with aging in both gen-

Eppur si muove: Formant Dynamics is Relevant for The Study of Speech Aging Effects

277

ders (Pellegrini et al., 2013; Albuquerque et al., 2014,

2019).

As far as we know, only Albuquerque et al. (2020)

indirectly studied the age effect in vowels’ formants

dynamics and showed that dynamic measurements of

F1-F3 result in higher rates of age group classiﬁcation

(senior/non-senior).

3 STATUS QUO: STATIC IS MORE

THAN ENOUGH

The vowel formant measurements have a long history

in the study of speech production, mainly because for-

mant descriptions are suited to articulatory interpreta-

tions of acoustic data and formant frequencies reﬂect

the length and conﬁguration of the vocal tract Mc-

Dougall and Nolan (2007); Kent and Vorperian (2018)

Previous studies have identiﬁed the ﬁrst two or

three formant frequencies as crucial acoustic corre-

lates for the identiﬁcation of vowels Peterson and

Barney (1952); Watson and Harrington (1999); Al-

murashi et al. (2019); Adank et al. (2004); Kent

and Vorperian (2018); Themistocleous (2017). Still,

vowel duration Almurashi et al. (2019); Albuquerque

et al. (2020) and fundamental frequency Zahorian and

Jagharghi (1993) as additional cues have been re-

ported to play a role in vowel discrimination.

The static approach has dominated acoustic anal-

ysis of vowels and is a convenient simpliﬁcation

because the vowel’s midpoint is the section of the

vowel that is least inﬂuenced by the contextual ef-

fects Watson and Harrington (1999); Peterson and

Barney (1952); Kent and Vorperian (2018); Hillen-

brand (2013) and also corresponds to the target po-

sition (i.e., vowel target), where a minimal shift in

formant frequencies is seen Harrington (2010); Al-

murashi et al. (2019); Van der Harst and Van de Velde

(2014).

Static vowel features may be sufﬁcient for vowel

classiﬁcation, as it has previously been shown for

some languages Williams et al. (2015); Sarvasy et al.

(2020); Almurashi et al. (2019).

Additionally, also cross-dialectal studies have re-

ported that vowel steady states are major acoustic

vowel features Ewald et al. (2017); Van der Harst and

Van de Velde (2014).

3.1 Vowel Dynamics is a Great

Challenge

Despite the fact that vowels show a signiﬁcant amount

of spectral movement throughout their course Hillen-

brand (2013); Williams and Escudero (2014); Sarvasy

et al. (2019), the analysis of the vowel dynamics is a

great challenge. For that, a solution has been to make

measurements at single time point Kent and Vorperian

(2018). The data obtained with this static approach

have the advantage of simplicity and economy Kent

and Vorperian (2018).

In a dynamic approach to extract vowel formants,

with a large number of time points, the formant

measures can wander in seemingly erratic directions

through the course of a vowel Thomas (2011). In ad-

dition, the plots to represent the vowels become more

difﬁcult to interpret, making the comparisons between

vowels challenging Thomas (2011).

Additionally, more complex models that allow to

analyze formant means, as well as the direction and

magnitude of the formant change, saturate on the

amount of speaker-speciﬁc information which can be

extracted and could begin ﬁtting noise Watson and

Harrington (1999); Ewald et al. (2017); Williams and

Escudero (2014).

4 ARGUMENTS FOR OUR

THESIS: DYNAMICS HAS

POTENTIAL

Having presented in the previous section our percep-

tion of the main counterclaims, we now present vari-

ous arguments to support our position.

Although the effectiveness of the formant frequen-

cies in vowel separation is indisputable, some studies

have also recognized that temporal information has

also been important for characterizing vowel qual-

ity Watson and Harrington (1999); Almurashi et al.

(2019); Williams and Escudero (2014). However, this

does not imply that duration is the best or even the

most relevant way to discriminate vowels Sandoval

and Utianski (2015). Whereas vowel duration is sen-

sitive to speaking rate, formant trajectory computed

relative to vowel duration is not Sandoval and Utian-

ski (2015).

Therefore, dynamic approaches based on formant

trajectories or combinations of measurement points

that sample the vowel formant pattern should be taken

into account Kent and Vorperian (2018).

Despite the fact that vowel formants may vary

to some extent according to phonetic context, some

formant movement may occur due to vowel inherent

spectral change (VISC) Williams et al. (2015), which

is deﬁned by Nearey and Assmann (1986) as the “rel-

atively slowly varying changes in formant frequen-

cies associated with vowels themselves, even in the

BIOSIGNALS 2021 - 14th International Conference on Bio-inspired Systems and Signal Processing

278

absence of consonantal context”.

4.1 Speech is Dynamic and is Affected

by Age

In speech production the jaw, tongue and other artic-

ulators move continuously in space and time through

several articulatory postures per second Rogers et al.

(2013). As a result, speech is inherently dynamic

Yuan (2013).

Dynamic features of speech offer greater scope for

variation among speakers, as they reﬂect the move-

ment of the individual’s speech organs, as well as

anatomical dimensions McDougall and Nolan (2007).

The explanations that have been advanced to ac-

count for age-related changes in vowel formant fre-

quencies have referred alterations in both dimensions,

mentioned above. Besides, for EP vowels, Albu-

querque et al. (2019) indicated that older speakers

might present speciﬁc articulatory adjustments dur-

ing speech. Therefore, a dynamic approach to study-

ing the age effect on vowel production might be im-

portant, since a static approach mainly demonstrates

anatomical differences among speakers.

Furthermore, the static measures do not address

information about how formant frequency changes in

time (Fox and Jacewicz, 2009).

4.2 Better Tools to go Beyond Snapshots

Analysis

For many years the methods and tools available for

researchers made only viable a small amount of mea-

sures Van der Harst and Van de Velde (2014), but all

changed with the availability of programmable tools

such as Praat, making the extraction of several mea-

sures over time as easy as extraction just one.

Meanwhile two sets of approaches to study the dy-

namic properties of vowels have been developed: a

series of successive time points (multiple time point

approach) and by curve-ﬁtting Van der Harst and Van

de Velde (2014).

In the multiple time point approach, measures

may involve comparing formant frequencies from

two or more discrete time points during vowel dura-

tion (e.g., two points Morrison (2013); Adank et al.

(2004); three points (i.e., the three-point model) Hil-

lenbrand et al. (1995); Almurashi et al. (2019)); ﬁve

points Fox and Jacewicz (2009); or thirty points

Williams and Escudero (2014)).

Alternatively, dynamic variations in the formants

F1 and F2 may be characterized by different measures

such as trajectory length (TL) and the spectral roc Fox

and Jacewicz (2009). Even though these latter metrics

incorporate more detailed spectral information, they

do not account for the directionality of the change

(i.e., if the frequencies actually increase or decrease

over time) Williams and Escudero (2014).

On the other hand, vowel dynamics can also be

expressed by curve-ﬁtting parameterizations, by ﬁt-

ting parametric curves such as polynomials Themis-

tocleous (2017) or discrete cosine transforms (DCTs)

Elvin et al. (2016); Williams and Escudero (2014);

Sarvasy et al. (2020); Watson and Harrington (1999)

to formant contours for quantifying the shape of com-

plex curve Brandt et al. (2018); Van der Harst and Van

de Velde (2014). These approaches allow to analyze

formant means as well as the direction and magnitude

of the formant change Ewald et al. (2017); Williams

and Escudero (2014). Furthermore, these avoid an ar-

bitrary choice of one or more vowel targets, which

can be tricky when a vowel appears either not having a

steady-state section, or if the formants reach a minima

or maxima at different times (Watson and Harrington,

1999).

Additionally, namely for recordings with poor

quality, formant analysis using averaged values over

multiple time points could be more reliable than only

using a single time point measurement Sarvasy et al.

(2020).

4.3 Dynamic Improve Classiﬁcation

Performance

A number of studies have reported that incorporating

measures of formant dynamics enhances the classiﬁ-

cation of vowels based on acoustic Almurashi et al.

(2019); Elvin et al. (2016); Yuan (2013); Jacewicz

et al. (2011b); Jacewicz and Fox (2013); Jacewicz

et al. (2009); Williams and Escudero (2014); Al-

Tamimi (2007); Chittaragi and Koolagudi (2019); Za-

horian and Jagharghi (1993). For instance, on Hijazi

Arabic (HA) vowels, Almurashi et al. (2019) revealed

that the static approach was sufﬁcient for vowel clas-

siﬁcation, but multiple time point approaches per-

formed better than a static approach Almurashi et al.

(2019). Nonetheless, dynamic acoustic properties for

vowel classiﬁcation has not presented the same im-

portance for all languages Sarvasy et al. (2020).

Williams and Escudero (2014); Elvin et al. (2016)

agree that in addition to formant trajectory means, du-

ration, magnitude and direction of formant trajectory

slope are essential acoustic parameters for represent-

ing the English vowels. Also, they concluded that

formant curvature (represented by the second DCT

coefﬁcients) was not necessary for classifying vow-

els (Elvin et al., 2016; Williams and Escudero, 2014),

but can aid with more ﬁne-grained/ subtle phonetic in-

Eppur si muove: Formant Dynamics is Relevant for The Study of Speech Aging Effects

279

formation from different speakers or different dialects

(Elvin et al., 2016).

Furthermore, formant dynamics plays a major role

in determining cross-dialectal acoustic differences for

some vowels Van der Harst and Van de Velde (2014);

Jacewicz and Fox (2013); Williams and Escudero

(2014). In Van der Harst and Van de Velde (2014)

both the multiple time point approach and the curve-

ﬁtting parameterization proved to be a clear improve-

ment on the static approach to describe regional vari-

ation in the dynamics of vowel formants of Standard

Dutch.

Formant dynamics are also useful for improving

the within-class separation of the Australian English

tense vowels from their lax counterparts Watson and

Harrington (1999).

In addition, spectral change patterns may provide

vowel phonetic details that are relevant in second-

language (L2) learning (e.g., Jin and Liu (2013)) and,

therefore, may prove to be useful for predicting L2

difﬁculties Elvin et al. (2016).

Although an approach only with two time points

was performed equally well in distinguishing vowel

categories as more sophisticated parametric curve

approaches Zahorian and Jagharghi (1993); Morri-

son (2013), a whole trajectory approach based on

parametric curves outperforms a two time points

approach for extract speaker information Morrison

(2013). Therefore, the measurements at more time

points pay off in return to more social information

than static measures. Van der Harst and Van de Velde

(2014)

4.4 Dynamics is Relevant to Vowel

Perception

The dynamics of formant-frequency patterns has been

reported to play an important role in vowel percep-

tion, mainly for English (Hillenbrand, 2013; Jin and

Liu, 2013; Jacewicz and Fox, 2012; Nearey and Ass-

mann, 1986; Strange et al., 1983; Strange, 1989).

Perceptual studies have shown that: vowel steady-

states can be removed with little or no effect on vowel

intelligibility; vowels with stationary spectral patterns

are not well identiﬁed; and also vowels in consonant

context are more accurately identiﬁed than isolated

vowels (Strange et al., 1983; Strange and Bohn, 1998;

Hillenbrand, 2013).

Therefore, the dynamic vowel formants provide

essential information about the characteristic of the

vowels (Strange et al., 1983), which support a dy-

namic speciﬁcation of vowel theory over static tar-

get theories (Strange and Jenkins, 2013). However,

Jacewicz and Fox (2012) demonstrated that formant

dynamics (i.e., VISC) can play a signiﬁcant role in

error identiﬁcation of some vowels more than others

by listeners.

5 CONCLUSION

The authors strongly believe in the potential of dy-

namic measurements of vowel formants for vowel

classiﬁcation. Also that these dynamic measurements

provide useful information about the speakers iden-

tity. As the vowel dynamics could be affected by age,

this information may be important to improve auto-

matic extraction of speaker age.

Summing up, the use of dynamic approaches as

a proxy for kinematic movement may be useful as a

means to track changes of the normal speech with age.

This could be validated further in experiments to de-

termine how formant trajectories change with age for

each gender.

As the population aging increases, the world is

facing new challenges He et al. (2016), and the au-

tomatic extraction from speech of speaker dependent

characteristics, such as age, has a wide range of ap-

plications that could be useful to improve the quality

of life of older people Sadjadi et al. (2016); Yue et al.

(2014).

6 FUTURE DIRECTIONS: A

PROPOSAL

In order to explore the impact of dynamic vowel for-

mants in EP vowel classiﬁcation throughout life span,

there is still a lot of work to be done, and the following

actions are required:

• to apply different dynamic approaches to exist-

ing speech corpora of EP (e.g., H

ainen et al.

(2012); Albuquerque et al. (2019)) in order to ob-

tain the dynamic vowel formants (i.e., F1, F2, and

F3) and to investigate which ones ﬁt better the

acoustic changes with age;

• to apply several classiﬁcation algorithms (e.g.,

Support Vector Machine (SVM) and (Deep) Neu-

ral Networks) to the same body of acoustic vowel

data, to investigate the performances of static and

dynamic information in vowel and age classiﬁca-

tion tasks in many ways:

– to explore the classiﬁcation of the EP vowel ac-

cording to their static and dynamic properties;

– to analyze percentage of errors in vowel clas-

siﬁcation of the EP vowels by age, in order to

BIOSIGNALS 2021 - 14th International Conference on Bio-inspired Systems and Signal Processing

280

analyze if there is an age group and/or vowels

more effectively classiﬁed based on dynamic

information compared with the static alone;

– to examine the age classiﬁcation performances

based on static information alone and on both

static and dynamic information of vowel for-

mants;

– to identify which are the dynamic cues with

more impact on age and vowel classiﬁcation

performance.

In our understanding, future work should seek to

determine the extent to which the dynamic aspects of

the vowel’s acoustic signal contribute to its identiﬁca-

tion beyond the static information which is available

at the vowel target, and whether these dynamic as-

pects of the vowels change with age. Also if there are

observable differences between genders.

ACKNOWLEDGEMENTS

This research was ﬁnancially supported by the project

Vox Senes POCI-01-0145-FEDER-03082 (funded by

FEDER, through COMPETE2020 - Programa Opera-

cional Competitividade e Internacionalizac¸

ao (POCI),

and by national funds (OE), through FCT/MCTES),

by the grant SFRH/BD/115381/2016 and by IEETA

(UIDB/00127/2020).

REFERENCES

Adank, P., Van Hout, R., and Smits, R. (2004). An acoustic

description of the vowels of Northern and Southern

Standard Dutch. J. Acoust. Soc. of Am., 116(3):1729–

1738.

Al-Tamimi, J. (2007). Static and dynamic cues in

vowel production: A cross dialectal study in Jorda-

nian and Moroccan Arabic. In 16th International

Congress of Phonetic Sciences (ICPhS), pages 541–

544, Saarbr

ucken, Germany.

Albuquerque, L., Oliveira, C., Teixeira, A., Sa-Couto, P.,

and Figueiredo, D. (2019). Age-related changes in

European Portuguese vowel acoustics. In INTER-

SPEECH, pages 3965–3969, Graz, Austria.

Albuquerque, L., Oliveira, C., Teixeira, A., Sa-Couto, P.,

Freitas, J., and Dias, M. S. M. (2014). Impact of age

in the production of European Portuguese vowels. In

INTERSPEECH, pages 940–944, Singapore.

Albuquerque, L., Teixeira, A., Oliveira, C., and Figueiredo,

D. (2020). The effect of dynamic acoustic cues on

age classiﬁcation. In SPPL2020: 2nd Workshop on

Speech Perception and Production across the Lifespan

(Poster), page 81.

Almurashi, W., Al-Tamimi, J., and Khattab, G. (2019).

Static and dynamic cues in vowel production in Hijazi

Arabic. In 19th ICPhS, pages 3468–3472, Newcastle.

Arias-Vergara, T., V

asquez-Correa, J. C., and Orozco-

Arroyave, J. R. (2017). Parkinson’s disease and aging:

analysis of their effect in phonation and articulation of

speech. Cognitive Computation, 9(6):731–748.

Barkana, B. D. and Zhou, J. (2015). A new pitch-range

based feature set for a speaker’s age and gender clas-

siﬁcation. Applied Acoustics, 98:52–61.

Benjamin, B. J. (1982). Phonological performance in geron-

tological speech. Journal of Psycholinguistic Re-

search, 11(2):159–167.

Brandt, E., Zimmerer, F., Andreeva, B., and M

obius, B.

(2018). Impact of prosodic structure and information

density on dynamic formant trajectories in German.

In 9th International Conference on Speech Prosody,

pages 119–123.

Braun, A. and Friebis, S. (2009). Phonetic cues to speaker

age: A longitudinal study. In Grewendorf, G. and

Rathert, M., editors, Formal Linguistics and Law,

pages 141–162. De Gruyter Mouton, Berlin.

Chittaragi, N. B. and Koolagudi, S. G. (2019). Acoustic-

phonetic feature based Kannada dialect identiﬁcation

from vowel sounds. International Journal of Speech

Technology, 22(3):1099–1113.

Decoster, F. and Debruyne, W. (1999). Acoustic differences

between sustained vowels perceived as young or old.

Log Phon Vocol, 24(1):1–5.

Eichhorn, J. T., Kent, R. D., Austin, D., and Vorperian,

H. K. (2018). Effects of Aging on Vocal Fundamental

Frequency and Vowel Formants in Men and Women.

Journal of Voice, 32(5):644.e1–644.e9.

Elvin, J., Williams, D., and Escudero, P. (2016). Dynamic

acoustic properties of monophthongs and diphthongs

in Western Sydney Australian English. J. Acoust. Soc.

Am., 140(1):576–581.

Ewald, O., Liina Asu, E., and Sch

otz, S. (2017). The for-

mant dynamics of long close vowels in three varieties

of Swedish. In INTERSPEECH, pages 1412–1416,

Stockholm, Sweden. ISCA.

Fletcher, A. R., McAuliffe, M. J., Lansford, K. L., and Liss,

J. M. (2015). The relationship between speech seg-

ment duration and vowel centralization in a group of

older speakers. J. Acoust. Soc. Am., 138(4):2132–

2139.

Fougeron, C., D’Alessandro, D., and Lancia, L. (2018). Re-

duced coarticulation and aging. J. Acoust. Soc. Am.,

144(3):1905.

Fox, R. A. and Jacewicz, E. (2009). Cross-dialectal varia-

tion in formant dynamics of American English vow-

els. J. Acoust. Soc. Am., 126(5):2603–2618.

ainen, A., Pinto, F. M., Dias, M. S., J

udice, A.,

Pires, C. G., Teixeira, V. D., Calado, A., and Braga,

D. (2012). The First European Portuguese Elderly

Speech Corpus. In IberSPEECH 2012: ”VII Jornadas

en Tecnolog

ıa del Habla” and “III Iberian SLTech”,

Madrid, Spain.

Harrington, J. (2010). Phonetic analysis of speech corpora.

John Wiley & Sons.

Eppur si muove: Formant Dynamics is Relevant for The Study of Speech Aging Effects

281

Harrington, J., Palethorpe, S., and Watson, C. I. (2007).

Age-related changes in fundamental frequency and

formants: a longitudinal study of four speakers. In

INTERSPEECH, pages 2753–2756, Belgium.

He, W., Goodkind, D., and Kowal, P. R. (2016). An ag-

ing world: 2015. International Population Reports,

P95/16-1.

Hillenbrand, J., Getty, L. A., Clark, M., and Wheeler, K.

(1995). Acoustic characteristics of American English

vowels. J. Acoust. Soc. of Am., 97(5 Pt 1):3099–3111.

Hillenbrand, J. M. (2013). Static and dynamic approaches

to vowel perception. In Morrison, G. S. and Assmann,

P. F., editors, Vowel inherent spectral change, pages

9–30. Springer.

Jacewicz, E. and Fox, R. A. (2012). The effects of cross-

generational and cross-dialectal variation on vowel

identiﬁcation and classiﬁcation. J. Acoust. Soc. Am.,

131(2):1413–1433.

Jacewicz, E. and Fox, R. A. (2013). Cross-dialectal differ-

ences in dynamic formant patterns in American En-

glish vowels. In Vowel inherent spectral change, pages

177–198. Springer.

Jacewicz, E., Fox, R. A., O’Neill, C., and Salmons, J.

(2009). Articulation rate across dialect, age, and gen-

der. Language variation and change, 21(02):233–256.

Jacewicz, E., Fox, R. A., and Salmons, J. (2011a). Cross-

generational vowel change in American English. Lan-

guage variation and change, 23(1):45–86.

Jacewicz, E., Fox, R. A., and Salmons, J. (2011b). Vowel

change across three age groups of speakers in three re-

gional varieties of American English. Journal of Pho-

netics, 39(4):683–693.

Jin, S.-H. and Liu, C. (2013). The vowel inherent spectral

change of English vowels spoken by native and non-

native speakers. J. Acoust. Soc. Am., 133(5):EL363–

EL369.

Johns III, M. M., Arviso, L. C., and Ramadan, F. (2011).

Challenges and opportunities in the management of

the aging voice. Otolaryngology - Head and Neck

Surgery, 145(1):1–6.

Kent, R. D. and Vorperian, H. K. (2018). Static Mea-

surements of Vowel Formant Frequencies and Band-

widths: A Review. Journal of Communication Disor-

ders, 74:74–97.

Lanitis, A. (2010). A survey of the effects of aging on bio-

metric identity veriﬁcation. International Journal of

Biometrics, 2(1):34–52.

Linville, S. E. (2001). Vocal aging. Singular Thomson

Learning, Australia, San Diego.

Linville, S. E. and Rens, J. (2001). Vocal Tract Resonance

Analysis of Aging Voice Using Long-Term Average

Spectra. Journal of Voice, 15(3):323–330.

Makiyama, K. and Hirano, S. (2017). Aging Voice.

McDougall, K. and Nolan, F. (2007). Discrimination of

speakers using the formant dynamics of/u:/in British

English. In International Congress of Phonetic Sci-

ences (ICPhS XVI), pages 1825–1828, Saarbr

ucken,

Germany.

Mertens, J., M

ucke, D., and Hermes, A. (2020). Aging

effects on prosodic marking in German: An acous-

tic analysis. In 2nd Workshop on Speech Perception

and Production across the Lifespan (Poster), London.

UCL.

Morrison, G. S. (2013). Vowel inherent spectral change

in forensic voice comparison. In Morrison, G. S.

and Assmann, P. F., editors, Vowel Inherent Spectral

Change, pages 263–282. Springer Berlin Heidelberg.

Nearey, T. M. and Assmann, P. F. (1986). Modeling the role

of inherent spectral change in vowel identiﬁcation. J.

Acoust. Soc. Am., 80(5):1297–1308.

Oliveira, C., Cunha, M. M., Silva, S., Teixeira, A., and

Sa-Couto, P. (2012). Acoustic analysis of European

Portuguese oral vowels produced by children. In

IberSPEECH, volume 328, pages 129–138, Madrid,

Spain.

Pellegrini, T., H

ainen, A., de Mare

uil, P. B., Tjalve, M.,

Trancoso, I., Candeias, S., Dias, M. S., and Braga, D.

(2013). A corpus-based study of elderly and young

speakers of European Portuguese: acoustic correlates

and their impact on speech recognition performance.

In INTERSPEECH, pages 852–856, Lyon.

Peterson, G. E. and Barney, H. L. (1952). Control methods

used in a study of the vowels. J. Acoust. Soc. of Am.,

24:175.

Qawaqneh, Z., Mallouh, A. A., and Barkana, B. D. (2017).

Deep neural network framework and transformed

MFCCs for speaker’s age and gender classiﬁcation.

Knowledge-Based Systems, 115:5–14.

Rastatter, M. P. and Jacques, R. D. (1990). Formant fre-

quency structure of the aging male and female vocal

tract. Folia phoniatrica, 42(6):312–319.

Rastatter, M. P., McGuire, R. A., Kalinowski, J., and Stu-

art, A. (1997). Formant frequency characteristics of

elderly speakers in contextual speech. Folia Phoni-

atrica et Logopaedica, 49(1):1–8.

Rogers, C. L., Glasbrenner, M. M., DeMasi, T. M., and

Bianchi, M. (2013). Vowel inherent spectral change

and the second-language learner. In Morrison, G. S.

and Assmann, P. F., editors, Vowel Inherent Spectral

Change, pages 231–259. Springer Berlin Heidelberg.

Sadjadi, S. O., Ganapathy, S., and Pelecanos, J. W. (2016).

Speaker age estimation on conversational telephone

speech using senone posterior based i-vectors. In In-

ternational Conference on Acoustics, Speech and Sig-

nal Processing (ICASSP), pages 5040–5044. IEEE.

Sandoval, S. and Utianski, R. L. (2015). Average Formant

Trajectories. Preprint submitted to Journal ofPhonet-

ics.

Sarvasy, H., Elvin, J., Li, W., and Escudero, P. (2019).

Vowel acoustic of Nungon, Papua New Guinea. In

ICPhS’19, pages 1714–1718, Melbourne, Australia.

Sarvasy, H., Elvin, J., Li, W., and Escudero, P. (2020). An

acoustic phonetic description of Nungon vowels. J.

Acoust. Soc. Am., 147(4):2891–2900.

Sataloff, R. T., Caputo Rosen, D., Hawkshaw, M., and

Spiegel, J. R. (1997). The aging adult voice. Jour-

nal of Voice, 11(2):156–160.

BIOSIGNALS 2021 - 14th International Conference on Bio-inspired Systems and Signal Processing

282

Sch

otz, S. (2006). Perception, analysis and synthesis of

speaker age, volume 47. Linguistics and Phonetics,

Lund University.

Sebastian, S., Babu, S., Oommen, N. E., and Ballraj, A.

(2012). Acoustic measurements of geriatric voice.

Journal of Laryngology and Voice, 2(2):81–84.

Statistics Portugal (2015). Envelhecimento da populac¸

residente em Portugal e na Uni

ao Europeia (Aging of

the resident population in Portugal and the European

Union). Destaque: informac¸

a comunicac¸

ao social.

Statistics Portugal (2019). Estimativas de Populac¸

ao Resi-

dente em Portugal - 2018 (Estimates of resident pop-

ulation in Portugal - 2018). Destaque: informac¸

comunicac¸

ao social.

Strange, W. (1989). Dynamic Speciﬁcation of Coarticulated

Vowels Spoken in Sentence Context. J. Acoust. Soc.

Am., 85(5):2135–2153.

Strange, W. and Bohn, O.-S. (1998). Dynamic speciﬁca-

tion of coarticulated German vowels: Perceptual and

acoustical studies. J. Acoust. Soc. Am., 104(1):488–

504.

Strange, W. and Jenkins, J. J. (2013). Dynamic speciﬁcation

of coarticulated vowels. In Morrison, G. S. and Ass-

mann, P. F., editors, Vowel Inherent Spectral Change,

pages 87–115. Springer Berlin Heidelberg.

Strange, W., Jenkins, J. J., and Johnson, T. L. (1983). Dy-

namic speciﬁcation of coarticulated vowels. J. Acoust.

Soc. Am., 74(3):695–705.

Themistocleous, C. (2017). Dialect classiﬁcation using

vowel acoustic parameters. Speech Communication,

92:13–22.

Thomas, E. (2011). Sociophonetics: An Introduction. Pal-

grave Macmillan.

Torre III, P. and Barlow, J. A. (2009). Age-related changes

in acoustic characteristics of adult speech. Journal of

Communication Disorders, 42:324–333.

Van der Harst, S. and Van de Velde, H. (2014). Variation

in Standard Dutch vowels: The impact of formant

measurement methods on identifying the speaker’s

regional origin. Language variation and change,

26(2):247–272.

Vipperla, R., Renals, S., and Frankel, J. (2010). Ageing

voices: The effect of changes in voice parameters on

ASR performance. J. Aud. Speech Music Process,

pages 1–10.

Watson, C. I. and Harrington, J. (1999). Acoustic evidence

for dynamic formant trajectories in Australian English

vowels. J. Acoust. Soc. Am., 106(1):458–468.

Watson, P. J. and Munson, B. (2007). A comparison of

vowel acoustics between older and younger adults. In

ICPhS XVI, pages 561–564, Saarbr

ucken.

Williams, D. and Escudero, P. (2014). A cross-dialectal

acoustic comparison of vowels in Northern and South-

ern British English. J. Acoust. Soc. Am., 136(5):2751–

2761.

Williams, D., Van Leussen, J.-W., and Escudero, P. (2015).

Beyond North American English: modelling vowel

inherent spectral change in British English and Dutch.

In 18th ICPhS, Glasgow.

World Health Organization (2012a). Ageing.

World Health Organization (2012b). Deﬁnition of an older

or elderly person.

Xue, S. A. and Hao, G. J. (2003). Changes in the Human

vocal tact due to aging and the acoustic correlates of

speech production: a pilot study. J Speech Lang Hear

Res, 46(3):689–701.

Yuan, J. (2013). The Spectral Dynamics of Vowels in Man-

darin Chinese. In INTERSPEECH, pages 1193–1197,

Lyon, France.

Yue, M., Chen, L., Zhang, J., and Liu, H. (2014). Speaker

age recognition based on isolated words by using

SVM. In CCIS2014, pages 282–286.

Zahorian, S. A. and Jagharghi, A. J. (1993). Spectral-

shape features versus formants as acoustic correlates

for vowels. J. Acoust. Soc. Am, 94(4):1966–1982.

Eppur si muove: Formant Dynamics is Relevant for The Study of Speech Aging Effects

283