Документ взят из кэша поисковой машины. Адрес оригинального документа : http://www.philol.msu.ru/~kedr/kedrova_tatry.pdf Дата изменения: Tue Aug 8 20:21:12 2006 Дата индексирования: Tue Oct 2 01:20:09 2012 Кодировка:

C

ONTRASTIVE STUDY OF THE MRI REPRESENTATION OF RUSSIAN VOWEL ARTICULATION (AGAINST FRENCH, GERMAN AND KOREAN ANALOGUES)

G. Ye. Kedrova a), L. M. Zakharov a), Yu. A. Pirogov b), N. V. Anis imo v
a) b)

b)

Philological Faculty, Moscow State Lomonosov University Moscow Educational-Research Center of Magnetic Tomography and Spectroscopy, Moscow State Lomonosov University Moscow, 119992 Moscow, GSP-2, Leninskije Gory, MGU, 1 build. of Humanities, Russia {kedr, leon}@philol.msu.ru; anisimovnv@mail.ru
Abstract: The MRI investigation of Russian vowels articulation inventory was realized upon original technology elaborated by the authors for real-time visualization of the speech articulation dynamics. One series of experiments dealt with articulation of the Russian sustained vowels. The data was collect ed in the course of several experimental sessions with the time gap of one months and one year. In all sessions the highest degree of MRI matching within various productions of a particular vowel by the same speaker was observed. Distinct inter-speaker differences were also analyzed and an appropriate interpretation suggested. The research focused on detailed visualization of lips and tongue evolutions, diversification of resonances in oral cavities, configuration of soft palate and larynx throughout the speech production process. Thereupon the Russian vowels articulation inventory was defined and confronted with corresponding data of the French, German and Korean. The factual identity of the Russian vowels articulation thus could be explicitly stated and appropriate instructions for the language teaching and learning elaborated.

1. Introduction
Speech articulation is a multidisciplinary area per se. Speech production is dir ected by social and individual psychological patter ns and regulations, basic physiological and neuropsychological mechanisms of mental processes and motor activity, anatomical constraints of huma n articulatory apparatus. Ther efore interdisciplinary investigation of basic la nguage phenomena ­ speech production and particularly, speech articulation, ­ was strongly r ecommended by prominent Russian linguists from the ver y beginning of the experimental linguistics and phonetics in Russia (I. Baudouin de Courtenay, V. Bogoroditsky, L. Scerba), and its r esults highly appreciated. Russian tradition of exploration of complex medical instrumental facilities (X-ray cine- and photo filming) in linguistic research dates from mid-1970s. A compr ehensive collection of Russian and Ukrainien visual data of selected speech articulation processes was hence accumulated by L. Skalozub and her disciples [1, 2]. Since the Xray investigation is a ver y harmful and extremely danger ous to the huma n hea lth empirical method, ma ny questions of that research work r emained unanswer ed, ther efor e numer ous significant aspects of the Russian speech articulation are still unexplored until nowada ys. However, a constant search for new less risky experimental instruments

never stopped in Russian linguistics as well. In this context wide range of facilities provided by MRI was consider ed as ver y promising, though unlike to the latest Western trends in experimental phonetics, the MRI investigation of speech production in Russia is definitely at its outset. Ther e are not so many authentic r eproductions of Russian articulatory patter ns traced upon on-line MR-imaging techniques, either for the vocal or consonant system of the Russian language. This situation is even mor e regr ettable as ther e is a considerable a mount of positive knowledge on articulation basis of other world languages (French, English, Ger man, Japanese, Kor ean, Swedish, etc.) available in the literature. Ther efor e our MRI investigation of ma in Russian vowel phonemes based upon a dmitted procedur es and techniques [3] (though expanded with several new original methods of the authors [4]) would b e an innovative one.

2. Materials and methods
Articulation patterns of ma in Russian vowel phonemes were investigated after having been traced upon articulation data collected from two refer ence subjects (ma le and female native speakers of Russian with standard pronunciation, free of a ny articulatory disease) using Magnetic Resonance Ima ging (MRI). MRI experiments wer e realized in Educational-Research Center of


Magnetic Tomography and Spectroscopy of the Moscow State Lomonosov University on a 1.5 T MR system (Tomikon S50 «Bruker »). The r eceiver coil was a quadrature neck coil (Fig. 1). Both speaking subjects wer e lying in supine position, any special mechanism of the hea d fixation not provided. They wer e r equired to produce a sustained phonation of Russian vowel phonemes: [a], [o], [u], [i], [e] several times during acquisitions of MR ima ges. The experimenter instructed the subject to initiate the speech process by counting ever y second, a couple seconds befor e the MRI acquisition starts. The mentioned above physical conditions of the MRI experiments (supine position, noisy environment, sustained pronunciation, etc.) raised a problem of the data validity according to the language investigation. An eva luation of effects of the supine position (a.k.a. "gravitational effect") in MR investigation of speech articulation was the ma in purpose of recently conducted research wor k [5], [6]. In both papers was concluded that the body position (supine or facing downwards) produces a certain affect on the position and shape of the tongue body, mostly perceptible on MRima ges of articulation process of the phoneme [a]. According to the authors, this impact was not perceptible to the ear or observable in a sonagramme, and ther efor e could be interpreted as a kind of compensation strategies of the spea ker, who tends to maintain acoustic consistency of his/her speech through adjustment of common articulation patterns to various spacial situations of speech production. The speech signal was simultaneously r ecorded through a microphone LifeVideoTM, fixed on a receiver's coil close to the speaker's mouth (see Fig. 1). Parallel r ecording of the starting points of MRI sequences was also prearranged. Both recordings wer e pr esented as a two-cha nnel oscillogram, which enables mor e pr ecise timing of MR ima ge to the particular phase of phonation in the future phoneme articulation identification. Apart from the MRI sessions, control audio a nd video r ecordings of the sa me speech data from the same refer ence subject wer e made in a professional record's studio environment. The total data set was collected in two separate experimental sessions with the time gap of one month and one year, but using the sa me r efer enc e subjects and with the sa me langua ge stimuli for all acquisitions.
Figure 1. Position of a reference subject in MRI experiment. To the right: two-channel oscillogram displaying phonation (upper line) and starting points of MRI sequences (bottom line).

The MR-ima ges wer e r econstructed and displayed in real time. Repetitions of vowel phonemes have been r ecorded for a male a nd a fema le speaker at respectively 2,0 and 2,7 ima ges per second. A refer ence subject has been reproducing each vowel phoneme up to 33 times in ever y experimental session, the aggregate total of releva nt MR-images being 768 items from the spea ker. All types of the data obtained in each experimental session are disposed on Figure 2.

Figure 2. Experimental data of MRI acquisitions' interface, post-MRI session control audio and video recordings incorporated.

2.1. MRI data
In our exper iments on MRI research of articulator y models of Russian vowel phonemes, MR scanning was executed on sagittal slice with the slic e thickness of 9 mm and to a field of view 200120 mm. Well known pulse sequence "gradient echo" was used with the following parameters: TR=12 ms, TE=5.5 ms, FA=10 degr ees. Under thes e conditions it was possible to obtain MR images with 2-2,7 frames in a second and with 3 mm inplane resolution. Numerous available through the literature MRI studies of sustained vowels in various languages provided us with articulatory contours of French


and Ger ma n vocal phonemes [i: e: a: o: u:], Korea n vowels [i: a: u:] outlined through experimental sessions with similar conditions [7, 8, 9, 10, 11].

2.2. Russian vowel phonemes' articulation contours.
The whole data set of MR ima ges collected in both experimental MRI acquisitions was identified a nd ascribed to each phase of a phoneme r ealization. In all exper imental sessions we've observed the highest degr ee of ima ge matching within each speaker's various perfor mances of a particular vowel that was repeated several times in differ ent vocal contexts. Thus, we consider our r esults as a

proof of the linguistic concept of phoneme defined as a "psychomotor complex for med in the early childhood via association of contiguity" [12]. However, evidence from the first and the second MRI acquisition sИa nces provided us with slightly differ ent MRI data - see Fig. 3. We r efer this misconsiliences between two experimental sessions of the shape and position of tongue body in the ma le-spea ker's pronunciation to previously mentioned "gravitational effect", mor e significa nt with the differ ent (slightly upset) position of his hea d. This effect was definitely less obvious in the fema le-speaker's articulations while the head position was mor e homogeneous.

Figure 3. Experimental data of MRI acquisitions, performed with the time gap of one year (upper two rows - male speaker, first (1) and second (2) experimental sessions; lower two rows ­ female speaker, first (3) and second (4) experimental sessions). Vowel phoneme sequence is (from left to right): [a: o: u: i: e:]

More distinct inter-spea ker differ ences also attached our attention a nd wer e thus analyzed in detail. Primarily, we would like to notice major discrepa ncies in the shape of back/front cavity volumes for the articulation of the vowel [a] between the two subjects. A perceptive analysis of

the appropriate r ecor dings detected distinct faucal timbr e of the phonation of [a] and apparently of [o] in fema le-speaker's pronunciation of experimental stimuli. We are inclined to charge the observed phenomena (first and for emost, noticeable significa nt decr ease in passage in the pharynx) on


the for going description of the gravitational effect. We suggest that in this case it should be interpreted as an action of the supine position, uncommon to the speaker. Thus, she has testified her inability to compensate with changed articulational patterns an impact of irregular spatial situation a nd of unusual environment.

3. Conslusion
In our investigation a set of repr esentative articulation contours for the Russian standard pronunciation was extracted upon a certain algorithm from the MR images taken during sustained phonation of five Russian vowels [a : o: u: i: e:]. We've supposed that the observed interspeaker a nd inter-session dissimilarity der ives from certain exter nal situational factors (such as gravitational effect or individual speaker's compensational strategies). Ther efor e, our research enabled elaboration of the ver ified articulation inventor y for basic Russian vowel phonemes. A comparative study of corresponding data referring to other languages (Ger man, French a nd Kor ean) showed mor e open character of front and mid-vowels in Russian (mor e substantial against Ger ma n vocal system a nd much less evident for the French articulation patter ns); marked shift backwards into pr e-faringea l zone, of general configurations of Korea n articulation base for the vocals (noticeable even on the contour of front vowel [i], while most evident contrast was observed for the Russian [a]); greater labialization and mor e close articulation of the Ger man back vowel [u] against its Russian counterpart; mor e shifted backwards articulation of the Russian vowel [u] in comparison to the French analogue; apparent center-oriented position of the Russian mixed vowels [e] and [o] compared to other languages. Thus, we are convinced that the comparative detailed study of articulation mechanisms in differ ent langua ges might form a solid basis for effective methods in for eign language teaching and learning.

Acknowledgements
This research was partly supported by a grant from the Interdisciplinary Research Foundation of Moscow State Lomonosov University. The authors are ver y grateful to collea gues fr om the Philological Faculty of MSU for collaborative a nd fruitful discussions.

References

[1] . . (1963): , . [2] . . (1979): ( ), . [3] Demolin, D., George, M., Lecuit, V., Metens, T., Soquet, A., Raeymaekers, H. (1997): Coarticulation and articulatory compensations studied by dynamic MRI, Proc Eurospeech97, 4346. [4] .., .., .., .. (2005): - . II , ­ 2005, 239240. [5] Engwall O. (2003): A revisit to the application of MRI to the analysis of speech production ­ testing our assumptions, Proc of 6th International Seminar on Speech Production, Sydney, Australia, Dec. 2003, 43-48. [6] Tiede M.K., Masaki S., Vatikiotis-Bateson E. Contrasts in speech articulation observed in sitting and supine conditions. Abstracts of the 5th Seminar on Speech Production: Models and Data. URL: http://www.phonetik.unimuenchen.de/Forschung/SPS5/abstracts/. [7] Demolin, D., Metens, T., and Soquet, A. (2000): Real time MRI and articulatory coordinations in vowels, Proceedings of the 5th Seminar on Speech Production: Models and Data, Kloster Seeon, Germany, 93-96 [8] Apostol L., Perrier P., Raybaudi M. & Segebarth C. (1999): 3D geometry of the vocal tract and interspeaker variability. Proceedings of the XIVth International Congress of Phonetic Sciences (ICPhS98), vol. 1, 443-446, San Francisco, California, USA. [9] KrЖger B.J., Winkler R., Mooshammer C., Pompino-Marschall B. (2000): Estimation of vocal tract area function from magnetic resonance imaging: Preliminary results. Proceedings of the 5th Seminar on Speech Production: Models and Data, Kloster Seeon, Germany, 333­336. [10] Engwall, O. and Badin, P. (2000): An MRI study of Swedish fricatives: coarticulatory effects, Proceedings of the 5th Seminar on Speech Production: Models and Data, Kloster Seeon, Germany, 297-300 [11] Yang Byunggon (1999): Measurement and synthesis of the vocal tract of Korean monophthongs by MRI, Proceedings of the XVth International Congress of Phonetic Sciences (ICPhS-99), San Francisco, USA. [12] .. (2004): . ., , ISBN 5-354-01021-7.