Документ взят из кэша поисковой машины. Адрес оригинального документа : http://www.philol.msu.ru/~kedr/kedrova-nasal.pdf Дата изменения: Mon May 4 01:56:02 2009 Дата индексирования: Tue Oct 2 01:28:02 2012 Кодировка:

MRI-based contrastive study of nasal and oral labial consonants' articulations in Russian

G. Kedrova, L. Zaharov Moscow State Lomonosov University Philological Faculty E-mail: kedr@philol.msu.ru Abstract
The MRI-based contrastive investigat ion of Russian nasal and oral hard and soft labial consonants' articulat ions was carried out upon original technology elaborated for real-time MRI visualization of the speech articulation dynamics. A crucial distinctive role of the velum configurations has been proved for this type of Russian consonantal production. Main differenc es between palatalize d and non-palatalized articulat ory patterns of the experimental consonants were depicted and some coarticulation constraints resultant from compatibility of various elements in CV clusters propounded.

N. Anisimov, Yu. Pirogov Moscow State Lomonosov University Research Centre for MRI and MRS

1. Introduction
Experimental investigation of Russian nasal articula tions has been until nowadays fragmentary and mainly based upon rather ambiguous interpre tation of acoustic data. The only direct observation of articulato ry processes in Russian speech dates back to the 1960s-1970s, when L. Skalozub, being a single Russian experimentalist in the field of direct observation of articul atory processes, completed a series of wide-ranging studies of articula tion dynamic models using x-ray processing of natural speech [1]. One of the main targets of her investigation of Russian articulat ion patterns was a comparison of articula tion dynamics of non-palatalized ("hard") and palatalized ("soft") consonantal pairs. The experimental data consisted of manually traced profiles of the articula tory tract

observed during phonation of various Russian consonants in certain frame sets (CVC, C'VC, etc.). One of the sidelines of her large-scale investigation of Russian consonantal articulation s was processing of articula tion profiles of palatal ized and nonpalatalized consonants of various places of articula tion and their potential clustering. Thus, she published some footnotes upon articulator y patterns of labial stops, nasal bilabial sonant [m] included. Skalozub hypothesized that all Russian hard labial and labiodental consonants ([v], [f], [m], [b], [p]) form a single class from the articulato ry point of view. She also stated that the vocal tract contours drawn upon x-ray filming of soft and hard consonants in experimental sessions of producing natural and pseudo speech stimuli differed drastical ly either in stationar y phase of a sound or in its dynamic deployment (motor stereotyping). Furthermore, these vocal tract configurations revealed a sort of commonalities in overall contour typical for all palatalized consonants independent of the place of articula tion, that L. Skalozub defined as dorsal articula tion posture in the pre-palatal section of the articula tory tract [2]. However, at that time a substantial insufficienc y in the data prevented L. Skalozub from more specific investigation of various types of consonants, therefore numerous significant aspects of the Russian speech articulati on processes ­ labial and nasal articulation s and coarticula tion processes among them ­ remain still unexplored. Our current research aims to fill this gap. The current research of nasal and non-nasal labial consonants' articulat ions forms an integral part of a broader magnetic resonance imaging investigation of


the complete inventory of articul atory motor patterns representative for contemporary Russian language pronunciation practices. Currently there is a considerable deficienc y of authentic experimental research of Russian articulato ry patterns based upon on-line MR-imaging techniques either for the vocal or consonant system of the Russian language. The first and unique precedent MRI-based research of the Russian vowels and consonantal articulato ry patterns, as well as pilot investigation of main anticipa tory coarticulati on models observed during anticipa tory and carry-over pausal fragments inbetween speech production activity (pre-adjustment and succeeding articulato ry movements for various vowels' production), has been recently reported in [3]. In this paper we expose for discussion some new results of a contrastive MRI investigation of Russian nasal and oral non-palatalized and palatalized labial consonant's articula tions (namely, comparison between articula tory patterns of the Russian hard phonemes [m], [b] and their soft counterparts [m'], [b']). The research in question differs from previous investigation of the Russian consonantal production processes both in technology (MRI vs. X-ray cinematography) and methodology (natural speaking vs. pre-regulated phonation), as well as in its local targets (multiple subjects, unbounded time frame vs. restrict ed number of speakers, limited time frame) and experimental material (random speech samples vs. targeted sampling).

2. MRI specifications methodology

and

experimental

MRI experimental techniques simultaneous audio recordings taken via a microphone fixed on a receiver's coil close to the speaker's mouth have been arranged. As this recording was strongly dominated by the MR scanning machine noise a parallel recording of the starting points of MRI sequences was also previewed. The recordings were delivered to two expert phoneticians as a twochannel oscillogra m along with recording of the same speech stimuli in a studio, enabling therefore precise timing of an MR image with a particul ar phase of phonation for future identificat ion and labeling. It's well known that any MRI investigation of major part of consonantal production and coarticulation processes is facing two serious challenges: 1) very short duration of the occlusive phase of a stop or affricate consonant and overall velocity of articulat ory movements, and 2) lack of MR signal from the teeth as objects consisting of solid calcifer ous tissue. The first problem could be mainly done by the on-line dynamic MRI technique, successfully applied in [4] and [5]. This technique relies on gated scanning of numerous repetit ions of the same speech sequence to reconstruct the real articula tory movement progressing in time. Eventually, the reconstructed sequence of on-line MR images could be plotted from across several repetiti ons. As we've had at our disposal for each subject his/her audio-recordings of the phonation processes consistent with time-markers of the starting points of MRI launchings, the exact temporal tagging of every MR image as well as its linguistic interpre tation became possible.

Technical parameters of MRI experimental sessions were as follows: MR scanning activit y was executed by the pulse sequence 'gradient echo' on sagittal cut with the slice thickness of 9 mm and to a field of view 200*120 mm. Under these conditions it has become possible to obtain MR images with 2.7 frames in a second and with 3 mm in-plane resoluti on. All MRI experiments were realized on a 0.5 T MR system (Tomikon S50 "Bruker") at the Research Center for MRI and MRS of Moscow State University. The speaking subject was lying in supine position with his head placed inside the receiver coil of the MRI unit. Consistently with generally accepted

3. Experimental material
According to the previously tested and approved experimental instructions a speaking subject was asked to produce a series of VCCV sequences (Russian pseudo-words with the second vowel stressed) containing Russian consonants [b], [m], [b'], [m'] in the [a]_[a], [i]_[i] and [a]_[e] vocalic frame set, that is a[b:] a, a[m:]a; a[b':]a, a[m':]a; i[b':]i, i[m':]i; a[b:]e, a[m:]e; a[b':]e, a[m':]e, repeating each stimulus several times during a session of MR image acquisition at voluntary pace. The main interest was focused on the [a]_[a] and


[a]_[e] contexts, being minimal distinctive positions for hard and soft consonantal correlates. Another contrastive pair has been formed by a nasal and nonnasalized (oral) bilabial consonants ([m], [b]) and their palatalized counterparts [m'], [b']). The whole data set of MR images of the experimental stimuli collecte d through all experimental MRI acquisitions consisted of 500 MR-images: at 100 images with [ab:А] and [am:А] articulati ons, [ab':А] and [am':А], [ab:И] and [am:И], [ab':И] and [am':И], [am':М] and [ab':М] ones. Each MR-image from experimental dataset has been identified and ascribed to the corresponding phase of every phoneme's realization.

Figure 3: MR-image of the occlusion phase of [b] with a manually traced contour (white line).

4. Results
Experimental data presented in our data set have a notable degree of articulato ry contours' matching for each type of hard and soft consonants under investigation irrespective of most vocalic contexts. So, one could suppose that the observed motor stereotyping presumably resulted from a phoneme's inherent properties and far less from a specific phonetic context. Typical vocal tract configuration patterns of hard and soft labial nasal consonants under investigation are exposed on figure 1 and 2.
Figure 4: MR-image of the occlusion phase of [b'] with a manually traced contour (white line).

A juxtaposition of articulato ry contours for phonemes [m] and [b] is exposed on figure 5.

Figure 5: MR-image of the occlusion phase of [m] with a contour of the [b] occlusion phase.

A juxtaposition of articulato ry contours for phonemes [m'] and [b'] is exposed on figure 6.
Figure 1: MR-image of the occlusion phase of [m] with a manually traced contour (white line).

Figure 6: MR-image of the occlusion phase of [b'] with a contour of the [m'] occlusion phase. Figure 2: MR-image of the occlusion phase of [m'] with a manually traced contour (white line).

5. Discussion
Our experimental data present a considerable grade of image matching within each class of hard

The same type of images of non-nasal consonants under investigation are exposed on figure 3 and 4.


and soft labial consonants, either nasal or non-nasal. Thus, the only difference s between nasal and nonnasal labial consonants' articul ations seem to be determined exclusively by specific configuration and position of the soft palate (velum) and the larynx (higher in palatalized nasals). Another noteworthy observation deals with apparent nasalization of the vowel [a] in the frame set with nasal consonants, as all the vowel's articulato ry contours in a data set of 100 items revealed certain degree of vowel's nasalization, indicated on an image by the fact that the passage to the nose cavity has not been fully closed. It is worth also mentioning that Russian phonological system does not have nasalized vowels as special phonemes. The opinion that the articula tion of vowel [a] in Russian is generally characterized by an incomplete closure of the velum (see also [6]) has not been confirmed in the experimental dataset of other speakers. The hypothesis on completely diverse articulator y patterns for majority of hard and soft consonants in Russian has been also strongly supported by our experimental data (see juxtaposition of consonants' contours on figure 7).

palatalized consonant phoneme within a CV cluster [m] + [e] or [b] + [e] is recommended as a standard pronunciation: [m'Иd'ium] (medium), [b'eZ] (beige), [b'ekСn] (bacon), etc.

5. Conclusion
The methodological and technological approach elaborated in current investigation of nasal and oral labial consonants in Russian has proved its validit y and could be recommended for implementation in the cases where other methods of direct on-line observation of articulato ry processes are not available. Preliminary observations on the physical nature of compatibility constraint s for various Russian articulatio ns would serve as a basis for further research in the field.

Acknowledgments
The authors sincerel y appreciate highly valuable comments of the anonymous reviewers, as well as permanent assistance of our colleagues from Philological Faculty and Faculty of Physics of Moscow State Lomonosov University. References
[1] Skalozub L. 1979. Dinamika zvukoobrazovanija (po dannym kinorentgenografirovanija). Naukova dumka, Kiev. [2] Skalozub L. 1963. Palatogrammy i rentgenogrammy soglasnyh fonem russkogo literaturnogo jazyka. Kiev. [3] Kedrova, G., Zaharov, L., Anisimov, N., Pirogov, Yu. 2008. Magnetic Resonance investigation of palatalized stop consonants and spirants in Russian, Pros. of the Int. Congress Acoustics'08 Paris, France, 2345-2350. [4] Demolin D., Metens T., Soquet A. 2000. Real time MRI and articulatory coordinations in vowels. In: Proc. of the 5th Seminar on Speech Production: Motor Planning and Articulatory Modelling, Kloster Seeon, Germany, 93-96. [5] Mathiak K., Klose, U., Ackermann, H., Hertrich, I., Kincses, W., Grodd, W. 2000. Stroboscopic articulography using fast magnetic resonance imaging. Int. J. of Language and Communication Disorders, 35 (3), 419-425. [6] Bondarko L. 1977. Zvukovoj stroj sovremennogo russkogo jazyka. Prosveshchenije, Moscow.

Figure 7: MR-image of the occlusion phase of [m] with the contour of the [m'] occlusion phase (left) and of [b] with the contour of the [b'] occlusion phase (right).

We suppose this difference s being the main reason for strong phonotactical constraints observed in the modern Russian language pronunciati on practice. The Russian standard pronunciation dictionar ies contain only very few words with the hard labial nasal consonant [m] or bi-labial stop consonant [b] preceding vowel [e] within the single syllable: [mer] (city mayor), [mejnstr'Мm] (mainstream), [ber] (rem), [bekvakАl] (back-vokal). It is worth mentioning that most of these words are recently adopted into Russian, or form a minimal distinctive pair (as for "city mayor"). In the absolute majority of other foreign adoptions of frequent occurrence in Russian a