Friction Sources Characterization for Fricative Consonants of Arabic

Fazia Karaoui and Amar Djéradi

Laboratoire de la Communication Parlée et de Traitement de Signal,

Université des Sciences et Technologies Houari Boumediene, Beb Ezzouar Alger, Algerie

Keywords: Fricatives Consonants, Vocal Tract Transfer Function, Noise Source Model.

Abstract: The objective of this work is the acoustic characterization of the friction source for Arabic voiced

consonants [v], [z], [ʒ], [h], [ς] and unvoiced ones [s], [f], [∫], [ħ], employing vocal tract transfer function

obtained from a direct measurement by the Pseudo Random Excitation of the human vocal tract and the

signal spectrum radiated at the lips. Assuming the separability of the source of the vocal tract considered as

a linear filter, the sources spectrum is obtained by the ratio of the output signal spectrum of the vocal tract

transfer function. The results are derived from data produced by two female and two male subjects.

1 INTRODUCTION

The fricatives are produced when a supra glottis

constriction is formed in the vocal tract; the air flow

through the constriction produces a jet (turbulence)

in the region downstream of the constriction (near

the walls of the vocal tract and / or the teeth that

form an obstacle). In addition to the turbulence, the

vocal cords vibrate in the case of voiced fricatives.

To characterize the Arabic fricative consonants our

study is divided in two parts: the first part is the

acoustic characterization of the vocal tract by

measuring the vocal tract transfer function using

Pseudo Random Excitation method. The second part

is determining the friction sources spectra involved

in the production of the studied Consonants.

2 MEASUREMENT OF THE

VOCAL TRACT TRANSFER

FUNCTION

The excitation of the vocal tract by a pseudo-random

binary sequence (PRE) is a technique for direct

measurement of the vocal tract transfer function in

human subjects developed by Djéradi et al. in 1991.

This method has several advantages over

previous methods for providing in this case a very

short measurement time (100ms), it is also

compatible with phonation (Djéradi et al., 1992).

2.1 Data Acquisition

The recordings were made in an ambient

environment. The fricative consonants [s, f, ∫, v, ʒ, z,

ħ, ς] were made by two female speakers of native

language Kabyle for the subject KF and Algerian

Arabic for the subject BS. The signal is digitized at

sampling frequency of 16 KHz. The measuring time

is about 100 ms, corresponding to a pseudo-random

sequence of 1024 points. Each consonant is recorded

ten times by each speaker

3 SOURCE SPECTRUMS

The pseudo random excitation method developed by

(Djéradi et al., 1991). Allows the measuring of the

vocal tract transfer function in phonation mode and

also the obtaining of the signal spectrum radiated at

the lips as shown in the diagram below:

Figure 1: Block diagram of PRE method.

Transfer function

Spectrum

Signal

Vocal trac

Speech

PRE

Intercorrelation

+

134

Karaoui F. and Djéradi A..

Friction Sources Characterization for Fricative Consonants of Arabic.

DOI: 10.5220/0004079901340136

In Proceedings of the International Conference on Signal Processing and Multimedia Applications and Wireless Information Networks and Systems

(SIGMAP-2012), pages 134-136

ISBN: 978-989-8565-25-9

 2012 SCITEPRESS (Science and Technology Publications, Lda.)

The sources spectrum is obtained by the ratio of

the output signal spectrum of the vocal tract transfer

function. The ten source spectrums of each

consonant are superposed in the same figure, and the

voiced consonants are opposed with the unvoiced

ones for the fourth speakers, in order to compare the

results, figure 2.

3.1 Analysis and Interpretation of

Results

The main observation is that the spectral curves

overlap very well and the peaks values obtained are

stable for the speakers in the corpus. For the dental

voiced/ unvoiced fricatives [z, s], we see the

emergence of two main peaks one at about 4500Hz

and the other at 3200 Hz, these resonances are

attributed to incisive that form the source of

obstacle. According to previous studies in particular

that of Shadle (Shadle, 1985). The value of these

peaks depend both of the shape of the obstacle, the

diameter of the constriction and the distance

between the constriction and the obstacle which may

explain the appearance of two peaks for the dental

source. The amplitude of the peak (at 4500Hz) of the

fricative [s] is 50dB and that of [z] is 35dB, the peak

at 3200Hz [s] is 40 dB and that of [z] is 30 dB. Note

that the amplitude spectrum for the dental voiced [z]

falls relative to that of the unvoiced [s], this is due to

the drop in pressure at the glottis. Note also that the

amplitude and slope of the spectrum differs between

subjects that are due to the pressure difference

generated by each subject.

For the post alveolar sources fricatives [∫, ʒ],

three main peaks appeared, the first is at 1500 Hz,

the second at 3200 Hz and the third is at 4500 Hz

representing the resonances of the vocal tract walls

and dental sources.

For the pharyngeal, the spectra obtained are

similar for the subjects. 4500Hz, 3200 Hz appeared

for the three fricatives [ς,

ħ, h], the band below 500

Hz is dominated by the effect of voicing for the

voiced fricatives [ς, h] for the fricative [ς] a broad

peak appeared at 400 Hz. For [h] is at 200 Hz. For

the [ħ] a broad peak appeared at 1500 Hz.

Figure 2: Sources spectra for unvoiced fricatives [s, ʃ ] (left) and voiced fricatives[z, ʒ] (right) spoken by subjects KF, BS.

FrictionSourcesCharacterizationforFricativeConsonantsofArabic

135

KF [ʕ]

KF [ħ]

KF [h]

Figure 3: Sources spectra for pharyngeal fricatives

consonants [ʕ, ħ, h] spoken by subject KF.

4 CONCLUSIONS

In this framework, we have implemented the PRE

method for direct measurement of the vocal tract

transfer function on human subjects.

From the transfer functions obtained and the

output signal spectrum we determined the acoustic

parameters of the friction sources concerned during

the production of the fricative consonants.

We conclude from this analysis that the spectra

of the sources of friction of voiced consonants and

unvoiced consonants are similar, they exhibit the

same resonances in the high frequencies, while in

the low band frequency the effect of the glottal

source is dominant. The amplitude spectrum for

voiced sources decreases compared to unvoiced ones

due to the drop in pressure at the glottis. The

amplitude also differs between speakers due to the

difference in pressure emitted by each subject during

the production of these fricatives.

These results are in agreement with the

preceding results studies particularly those obtained

by Abeer Alwan and Shrikanth Narayanan for [s, ʃ].

On prospects, continue this study and propose a

complete model for a source of friction for Arabic

consonants:

- A signal generator whose spectrum is the one we

obtained in this work

- Command parameters that would allow the use

of this generator in a synthesis of fricative

consonants.

REFERENCES

Djéradi A., Badin P. & Guérin B., (1992). Effets de

couplage subglottique: mesure et modélisation dans le

domaine fréquentiel pour les fricatives d'arrière de

l'arabe. In 19èmes Journées d'Etude sur la Parole, pp.

13-18. Bruxelles, Belgique, 1992.

Christine Helen Shadle. The Acoustics of Fricative

Consonants March, 1985.

Shrikanth Narayanan, AbeerAlwen. Noise Source Models

for Fricative Consonants. IEEE transactions on speech

and audio processing, vol, 8, no.2,March 2000.

Badin P., (1991). Fricative consonants: acoustic and X-ray

measurements. Journal of Phonetics, 19, 397-408.

Djéradi A., Guérin B., Badin, P. & Perrier, P., (1991).

Measurement of the acoustic transfer function of the

vocal tract: a fast and accurate method. Journal of

Phonetics, 19, 387-395.

Beautemps D., Badin P. & Laboissière R., (1993).

Recovery of vocal tract midsagittal and area function

from speech signal for vowels and fricative

consonants. In 3rd EuroSpeech Conference, vol. 1, pp.

73-76. Berlin, Germany, September 1993.

Guérin B., Badin P., Grabbe-Georges, E., Moulinier, A.,

Shadle C. H., Scully C. & Tzavali E., (1991). Mesure,

caractérisation et modélisation des sons fricatifs,

Quatrième rapport semestriel du contrat SCIENCE

SC1*147-C.

SIGMAP2012-InternationalConferenceonSignalProcessingandMultimediaApplications

136