Electromagnetic Midsagittal Articulometer Research Articles

Schwas with and without active gestural control

This study investigates whether some epenthetic vowels are targetless, that is, whether they can arise from the timing of the surrounding consonants [C. P. Browman and L. Goldstein, Papers in laboratory phonology II: Gesture, segment, prosody, 26–56 (1992)]. Specifically, the difference in targetlessness between past tense and lexical schwas in American English is examined. Articulatory data were collected from three speakers of American English using an electromagnetic midsagittal articulometer. The stimuli included phrases with past tense and lexical schwas embedded in a common environment: for example, ‘‘If cheated even once’’ (past tense schwa) and ‘‘If Cheetah’d even known’’ (lexical schwa). If the tongue body is assumed to be controlled continuously by the targetful vowels, the tongue body position ‘‘during a schwa’’ should not differ significantly from the tongue body position during the preceding vowel in the epenthetic schwa tokens while it would in the lexical schwa tokens. In addition, if the lexical schwas have tongue body gestures associated with them but the past tense schwas do not, a significant interaction between schwa type and vowel context is expected. These results would provide evidence of a difference in targetlessness between the past tense and lexical schwas in American English. [Work supported by NIH.]

Acoustic and articulatory differences across word position for American English /r/

Typically, American English /r/ is manifested with different formant frequency patterns in word-initial, word-final and word-medial position. In particular, it has been reported (Lehiste, 1962; Espy-Wilson, 1992) that initial /r/’s show lower third formants. This may be due to differences in position or configuration of the primary articulator for /r/—the tongue—or it may reflect an increased degree of lip-rounding. Differences in tongue shape across word position have been reported (Zawadaski and Kuehn, 1982), but it is not clear how reported differences may relate to formant lowering. In this study, eight native speakers of American English recorded words with initial, final and medial /r/’s. Articulatory movements of the tongue tip, tongue blade and tongue dorsum, lips and jaw were tracked simultaneously by Electro-Magnetic Midsagittal Articulometer (EMMA). For each speaker, we compare spatial position, time course, and trajectory for articulators across positional variants. Preliminary data suggest that tongue articulators show similar time course in all three positions. Results are discussed with regard to acoustical modeling studies, theories of articulatory strengthening and interarticulator timing. [Research supported by NIH and NSF.]

The relevance of F4 in distinguishing between different articulatory configurations of American English /r/

American English /r/ is often cited as a segment exemplifying a many-to-one articulatory to acoustic relationship. Articulatory data show a continuum of vocal tract configurations for /r/, ranging from ‘‘retroflexed’’ to ‘‘bunched,’’ all involving three constrictions: (1) in the pharynx, (2) along the palate, and (3) at the lips. Researchers have noted, however, that all of these configurations result in similar acoustic profiles, at least for F1, F2, and F3. In this study, acoustic and articulatory data were used to investigate the mechanisms which determine the shape of the F4 trajectory for /r/. Articulatory positioning data and simultaneous acoustic data were collected from five subjects who said the phrase ‘‘Say warav for me.’’ The Electromagnetic Midsagittal Articulometer (EMMA) system was used to record the positions of three points on the tongue. Preliminary analysis based on the y position of the front, mid, and back transducers suggests that the behavior of F4 changes with different /r/ configurations and that the frequency of F4 is based on the length of the cavity between the palatal and pharyngeal constrictions.

Articulatory trade-offs reduce acoustic variability in American English /r/ productions

Acoustic and articulatory recordings indicate that speakers utilize trade-offs between different articulatory parameters to maintain acoustic stability during American English /r/ productions. Electromagnetic midsagittal articulometer data from seven subjects producing /r/ in five phonetic contexts were analyzed. Articulator configurations for /r/ in different contexts showed systematic trade-offs between constriction length, constriction area, and front cavity volume. The hypothesized effects of these three parameters on F3, derived from acoustic theory, were verified using a 51-tube model of the vocal tract derived from structural MRI scans of a speaker producing /r/. All seven subjects utilized a trade-off between front cavity volume and the length of the constriction formed by the tongue. Further analysis of acoustic and articulatory variabilities indicates that the net effect of these articulatory covariances is a considerable reduction in acoustic variability, thus allowing large contextual variations in vocal tract shape; these contextual variations in turn appear to reduce the total amount of articulatory movement required. These findings contrast with the view that speaking involves a target set of vocal tract constrictions for each phoneme and provide strong evidence for acoustic theories of speech production. [Work supported by NIDCD, NSF, and the Sloan Foundation.]

A maximum likelihood approach to estimating speech articulator positions from speech acoustics

An algorithm called maximum likelihood continuity mapping (MALCOM) will be presented. MALCOM recovers the positions of the tongue, jaw, and lips from measurements of the sound-pressure waveform of speech. Unlike other techniques for recovering articulator positions from speech, MALCOM does not require training on measured or modeled articulator positions, and MALCOM does not rely on any particular model of sound propagation through the vocal tract. The algorithm categorizes short-time windows of speech into a finite number of sound types, and assumes the probability of using any articulator position to produce a given sound type can be described by a parametrized probability density function. MALCOM uses maximum likelihood estimation techniques to: (1) find the most likely smooth articulator path given a speech sample and a set of probability density functions (one density function for each sound type); and (2) change the parameters of the probability density functions to better account for the data. The data set used for evaluating the continuity mapping algorithm is comprised of simultaneously collected articulator and acoustic measurements made using an electromagnetic midsagittal articulometer on a human subject. Comparisons between measured articulator positions and those recovered using continuity mapping will be presented.

English /r/ and /l/ production in American and bilingual Japanese subjects

The tongue shapes of American and bilingual Japanese subjects were evaluated when producing words that include /r/’s and /l/’s in English. An electromagnetic midsagittal articulometer (EMMA) system was utilized to observe and record the tongue shape. A total of six transducer coils was put on the tongue. Four were on the superior side of the tongue, and two were on the inferior side. The tongue shapes of the American subjects when producing /r/’s generally agreed with the earlier studies done by Delatree and Freeman [Linguistics 44, 28–68 (1968)]. On the other hand, the Japanese subjects had more uniform tongue shapes across subjects, which was thought to correlate with the Japanese /r/’s. [Work supported by NIH.]

Evaluating a topographical mapping from speech acoustics to tongue positions

The continuity mapping algorithm—a procedure for learning to recover the relative positions of the articulators from speech signals—is evaluated using human speech data. The advantage of continuity mapping is that it is an unsupervised algorithm; that is, it can potentially be trained to make a mapping from speech acoustics to speech articulation without articulator measurements. The procedure starts by vector quantizing short windows of a speech signal so that each window is represented (encoded) by a single number. Next, multidimensional scaling is used to map quantization codes that were temporally close in the encoded speech to nearby points in a continuity map. Since speech sounds produced sufficiently close together in time must have been produced by similar articulator configurations, and speech sounds produced close together in time are close to each other in the continuity map, sounds produced by similar articulator positions should be mapped to similar positions in the continuity map. The data set used for evaluating the continuity mapping algorithm is comprised of simultaneously collected articulator and acoustic measurements made using an electromagnetic midsagittal articulometer on a human subject. Comparisons between measured articulator positions and those recovered using continuity mapping will be presented.

Speaking rate and associated kinematic patterns of tongue movements

Kinematically based measures of effort for tongue movements were examined for a speaker with a cochlear implant. Turning the speech processor of the implant on- and off-induced changes in the subject’s speaking rate. An electromagnetic midsagittal articulometer (EMMA) was used to track movements of the tongue body toward and away from the steady-state target for the /u/ vowel in the utterance ‘‘...who hid...’’. Although the tongue body movements toward the /u/ were quite small—on the order of a millimeter—kinematic measurements for those movements still exhibited relationships that were strikingly similar to results of past findings, which have shown strong relations between movement distance and duration. Similarly, values for velocity profile shape (the ratio of peak velocity to average velocity) appeared to follow increases in movement duration. According to the measures of peak velocity and energy, the most ‘‘effort’’ for these small movements was exerted in the condition when the cochlear implant had been off for the longest time. Examination of the number of velocity peaks and of tongue body trajectories for each utterance revealed that ‘‘smoother’’ and less effortful movements appeared with greater frequency when the cochlear implant was on. [Work supported by N.I.D.C.D.]

Underlying control strategies for the production of stop consonants

The underlying control strategies utilized in the production of stop consonants were investigated using an electromagnetic midsagittal articulometer [J. Perkell etal., J. Acoust. Soc. Am. 92, 3078–3096 (1992)]. The movements in the midsagittal plane of points on the lower jaw, lips, tongue blade, and tongue body were measured for three normal hearing, normal speaking subjects during the production of single-syllable words composed of a stop or nasal consonant followed by the vowel /ɑ / or /i/ and the consonant /t/, and spoken in a carrier phrase. Half the tokens were preceded by the fricative consonant /s/. The data show that the maximum downward velocity of the lower jaw occurs approximately at the time of the consonant release. For /sCV/ syllables, the lower jaw is held in a high position to assist in noise generation for /s/, and consequently the movement for a following labial or velar consonant follows a time course that is different from that of /CV/ syllables. The movement of the articulators during the production of these utterances at a normal speaking rate appears to have a bandwidth limitation of 5–7 Hz. [Work supported in part by grants from NIDCD.]

Inferring articulator positions from acoustics: An electromagnetic midsagittal articulometer experiment

To examine the mapping from acoustics to articulation, simultaneous articulatory and acoustic measurements were made of 90 vowel‐to‐vowel transitions produced by a Swedish speaker. Articulator positions were measured using an electromagnetic midsagittal articulometer that tracked seven receiver coils placed on the lips, jaw, and tongue. The vocal tract transfer function associated with each articulator position was estimated using cepstral analysis on short windows (25.6 ms) of the acoustic signal. The transfer functions were then vector quantized, giving each articulator configuration a corresponding vector quantization code. Given an acoustic signal and its corresponding vector quantization code (code 1 for example), an estimate of the position of any coil during the production of a speech signal can be made by averaging all the positions the coil assumed during the production of sounds labeled with vector quantization code 1. This produces a codebook that allows articulator positions to be estimated from acoustic signals. On a data set not used for training, correlations between estimated and actual receiver coil positions vary strongly with the place of attachment, but correlations above 0.9 are common for coils on the tongue. [Work supported by NIH.]

Trading relations between tongue-body raising and lip rounding in production of the vowel /u/: a pilot "motor equivalence" study.

Articulatory and acoustic data were used to explore the following hypothesis for the vowel /u/: The objective of articulatory movements is an acoustic goal; varying and reciprocal contributions of different articulators may help to constrain acoustic variation in achieving the goal. Previous articulatory studies of similar hypotheses, expressed entirely in articulatory terms, have been confounded by interdependencies of the variables being studied (e.g., lip and mandible displacements). One case in which this problem may be minimized is that of lip rounding and tongue-body raising (formation of a velo-palatal constriction) for the vowel /u/. Lip rounding and tongue-body raising should have similar acoustic effects for /u/, mainly to lower F2. In multiple repetitions, reciprocal contributions of lip rounding and tongue-body raising could help limit F2 variability for /u/; thus this experiment looked for complementary covariation (negative correlations) in measures of these two parameters. An electro-magnetic midsagittal articulometer (EMMA) was used to track movements of midsagittal points on the tongue body, upper and lower lips, and mandible for large numbers of repetitions of utterances containing /u/. (Interpretation of the data was aided by results from area-function-to-formant modeling.) Three of four subjects showed weak negative correlations, tentatively supporting the hypothesis; a fourth showed the opposite pattern: positive correlations of lip rounding and tongue raising. The results are discussed with respect to ideas about motor equivalence, the nature of speech motor programming, and potential improvements to the paradigm.

Measuring articulatory movements with an electro-magnetic midsagittal articulometer (EMMA) system

It is difficult to find experimental evidence for underlying mechanisms in speech production, largely because speech acoustics and kinematics are very variable, within and across speakers. Thus one approach to exploring speech production is with statistical analyses of large quantities of articulatory and acoustic data, gathered in a systematic and efficient fashion. If used carefully, EMMA systems can transduce large amounts of accurate movement data simultaneously from midsagittal points on the maxilla, mandible, tongue, lips, and possibly the velum. The use of such a system in combination with appropriate analysis techniques can make it possible to investigate speech kinematics while taking variability into account. This presentation will cover a number of aspects of the use of an EMMA system including: the theory of operation, subject safety, calibrations, preparing for and running an experiment, signal processing, data extraction, checking and using the data, and limitations of this kind of data. [Work supported by N.I.D.C.D.]

Electromagnetic midsagittal articulometer systems for transducing speech articulatory movements

This paper describes two electromagnetic midsagittal articulometer (EMMA) systems that were developed for transducing articulatory movements during speech production. Alternating magnetic fields are generated by transmitter coils that are mounted in an assembly that fits on the head of a speaker. The fields induce alternating voltages in a number of small transducer coils that are attached to articulators in the midline plane, inside and outside the vocal tract. The transducers are connected by fine lead wires to receiver electronics whose output voltages are processed to yield measures of transducer locations as a function of time. Measurement error can arise with this method, because as the articulators move and change shape, the transducers can undergo a varying amount of rotational misalignment with respect to the transmitter axes; both systems are designed to correct for transducer misalignment. For this purpose, one system uses two transmitters and biaxial transducers; the other uses three transmitters and single-axis transducers. The systems have been compared with one another in terms of their performance, human subjects compatibility, and ease of use. Both systems can produce useful midsagittal-plane data on articular movement, and each one has a specific set of advantages and limitations. (Two commercially available systems are also described briefly for comparison purposes). If appropriate experimental controls are used, the three-transmitter system is preferable for practical reasons.

Tongue surface deformation during obstruent stop consonants

When an obstruent stop consonant is produced, there is an increase in intraoral pressure, and this increased pressure can result in outward movement of the vocal tract walls. Following release of the consonant, the intraoral pressure decreases and the walls are expected to move inward with a time constant that depends on the physical properties of the surfaces. An electromagnetic midsagittal articulometer (EMMA) was used to measure the vertical position of the tongue dorsum during a number of repetitions of five utterances: /ama/, /apa/, /aba/, /ampa/, and /amba/. Results seem to confirm the physical description given above: the average decrease in tongue dorsum position from implosion to release of the consonant was 0.4 mm for /m/, 2.7 mm for /b/, and 1.3 mm for /p/. The smaller movement for /p/ than for /b/ may be due to active stiffening of the tongue during /p/, partly counteracting the effect due to intraoral pressure. These displacements and their rates of movement are consistent with published data on average subglottal pressure, and compliance and resistance of vocal tract surfaces. These data have been used to refine current estimates of physical parameters of articulator surfaces during speech production. [Work supported by NIH Grant No. DC00075.]

An experimental analysis of spirantization in Spanish

An experimental analysis is presented of the spirantization phenomenon in Spanish. Spirantization is a phonological rule of all dialects of Spanish by which underlying voiced obstruents /b/, /d/, and /g/ surface as voiced continuants /β/, /edh/, and /ṽ/ when preceded by other continuants. It is argued that spirantization can be explained as the result of the overlapping and blending of articulatory gestures. Gestures are specified in two main dimensions: constriction location and constriction degree [C. B. Browman and L. Goldstein, Phonology 6, 201–251 (1989)]. The overlap of underlying obstruents with the preceding continuants results in blending of the two constriction degrees, thus creating another continuant. An electromagnetic midsagittal articulometer was used to record movement of the jaw, lower and upper lip, tongue tip, tongue blade, and tongue body, together with the acoustic signal, during the production of /b/, /d/, and /g/ in a number of phonological contexts. Results show that the magnitude of the obstruent gesture is indeed affected by that of the preceding continuant in terms of duration, constriction degree, and articulator velocity, and that the relationship between gestures is consistent across points of articulation. [Work supported by NIH Grant Nos. DC-00121 and HD-01994 to Haskins Laboratories.]