Articulatory Positions Research Articles

Estimation of Place of Articulation During Stop Closures of Vowel–Consonant–Vowel Utterances

Production of vowel-oral stop consonant-vowel utterances involves movement of articulators from the articulatory position of the initial vowel towards that of the oral stop closure, and then to that of the final vowel. As the closure segments have zero or low signal energy, linear predictive coding (LPC)-based estimation of vocal tract shape fails during stop closure. This paper reports a technique for estimation of place of articulation during stop closures by performing bivariate polynomial modeling on vocal tract area values during transition segments preceding and following the closure. The technique with second-degree polynomial modeling was found to be suitable for estimating the place of maximum constriction during stop closure segments of vowel-consonant-vowel utterances with bilabial, alveolar, and velar stops. The estimated places compared well with the actual constriction locations observed from the articulatory data. The technique may be useful for improving effectiveness of speech-training aids for production of stop consonants by providing visual feedback of the estimated place of articulation.

Understanding VOT Variation in Spontaneous Speech

This paper reports a corpus study on the variation of VOT in voiceless stops in spontaneous speech. Two speakers’ data from the Buckeye corpus are used: one is an older female speaker with a low speaking rate while the other is a younger male speaker with an extremely high speaking rate. Linear regression analysis shows that place of articulation, word frequency, phonetic context, speech rate and utterance position all have an effect on the length of VOT. However, altogether less than 20% of the variation is explained in both speakers, which suggests that pronunciation variation in spontaneous speech is a highly complicated phenomenon which might need more sophisticated modeling. Our results also show a great deal of individual differences.

Cleft Palate and/or Velopharyngeal Dysfunction: Assessment and Treatment

Abstract Children with cleft palate and/or velopharyngeal dysfunction often exhibit complicated clinical findings. In order to differentiate a resonance disorder from an articulatory/phonological disorder, the speech-language pathologist must be able to evaluate the function of the velopharyngeal mechanism and have knowledge of aberrant articulatory postures that create compensatory articulations. This paper will address current assessment techniques and treatment strategies for this clinical population. Methods for collaborating with the transdisciplinary cleft palate/craniofacial team will be provided.

Accuracy Assessment for AG500, Electromagnetic Articulograph

The goal of this article was to evaluate the accuracy and reliability of the AG500 (Carstens Medizinelectronik, Lenglern, Germany), an electromagnetic device developed recently to register articulatory movements in three dimensions. This technology seems to have unprecedented capabilities to provide rich information about time-varying positions of articulators. However, strengths and weaknesses of the system need to be better understood before the device is used for speech research. Evaluations of the sensor positions over time were obtained during (a) movements of the calibration device, (b) manual movements of sensors in a cartridge within the recording field of the cube, and (c) various speech tasks. Results showed a median error to be under 0.5 mm across different types of recordings. The maximum error often ranged between 1 and 2 mm. The magnitude of error depended somewhat on the task but largely on the location of the sensors within the recording region of the cube. The performance of the system was judged as adequate for speech movement acquisition, provided that specific steps are taken for minimizing error during recording and for validating the quality of recorded data.

Distinguishing Place of Consonant Articulation using the Aurora System

The goal of this project is to determine the effect of bite‐block and speaking rate manipulation on lingual targets for consonants using a recently developed Aurora, a 3‐D electromagnetic movement tracking system (NDI). Bite‐block and speaking rate manipulations are common intervention techniques with speakers with dysarthria and apraxia of speech. Therefore, understanding their effect on consonant target regions is essential for predicting the outcomes of articulatory training in these clinical populations. In this project, articulatory positions of two sensors on the tongue tip and dorsum will be recorded independent of the head. Stops /t, k/ and fricatives /s, sh/ will be embedded in aCa syllables. A bite block condition will be used to eliminate jaw contribution to tongue movements. The location and size of articulatory regions associated with each consonant will be compared in bite block and no‐block conditions. Additionally, the rate effect on the location and size of the consonant articulatory regions will be examined. Potential clinical implication of the finding on articulatory intervention will be considered.

Compensatory responses to unexpected jaw loading during speech observed with magnetometry

Observations were made of articulator and formant trajectories during speech perturbed by unexpected mechanical loads applied to the jaw. Subjects produced multiple repetitions of the phrase ‘‘see X avis,’’ where X was one of ‘‘raw,’’ ‘‘rob’’ or ‘‘rod.’’ Perturbation forces were applied using a jaw-coupled robot, and triggered using a thresholding criterion applied to real-time tracking of the initial low to high transition in the first formant. Perturbations were delivered one out of every five repetitions, selected at random, with half applied upwards and half downwards, and forces sustained throughout the remainder of the utterance. Three EMMA sensors located on the anterior midsagittal surface of the tongue in addition to lower mandibular and upper lip sensors were used to track articulator positions. Audio and surface EMG (masseter, ABD, and OOS) were collected concurrently. Formants show initial deviation from control trajectories and then recovery that begins approximately 75 ms after the onset of perturbation. Compensation in most instances is nearly complete, even though jaw position does not recover its unperturbed trajectory. Perturbed tongue sensor trajectories are strongly distinguished by coda type, but show a pattern converging on the unperturbed tongue shape consistent with the tongue’s presumed role in effecting formant recovery.

Jaw and Order

It is well-accepted that the jaw plays an active role in influencing vowel height. The general aim of the current study is to further investigate the extent to which the jaw is active in producing consonantal distinctions, with specific focus on coronal consonants. Therefore, tongue tip and jaw positions are compared for the German coronal consonants /s, f, t, d, n, l/, that is, consonants having the same active articulators (apical/laminal) but differing in manner of articulation. In order to test the stability of articulatory positions for each of these coronal consonants, a natural perturbation paradigm was introduced by recording two levels of vocal effort: comfortable, and loud without shouting. Tongue and jaw movements of five speakers of German were recorded by means of EMMA during /aCa/ sequences. By analyzing the tongue tip and jaw positions and their spatial variability we found that (1) the jaw's contribution to these consonants varies with manner of articulation, and (2) for all coronal consonants the positions are stable across loudness conditions except for those of the nasal. Results are discussed with respect to the tasks of the jaw, and the possible articulatory adjustments that may accompany louder speech.

Spatial-temporal patterns of vowel articulation: A cross-speaker study

An important goal in speech production research is to establish a quantitative description of normal or typical articulatory behavior. This description should be expressed in a compact but meaningful way and be robust enough to generalize across a large number of speakers who differ in physical size, gender, age, etc. This is a significant challenge given the numerous degrees of freedom within the articulatory system and the range of speaker variability. However, Hashi et al. [J. Acoust. Soc. Am. 104, 2426–2437 (1998)] used x-ray microbeam data to develop a procedure for normalizing vocal tract size and shape. This procedure reduced cross-speaker variation in the articulatory postures of vowels. The authors limited their analysis to a single time slice at the acoustic steady state. There is little available information regarding speaker-general time-varying articulatory patterns associated with vowel production. This study will address this issue. Fifty-three healthy speakers from the University of Wisconsin x-ray microbeam database served as subjects. The speech task consisted of productions of American English vowels in a /sVd/ context. Analysis focused on the movement of a number of vocal tract markers over the duration of each vowel. Vowel-specific spatial temporal patterns common to the speaker group will be described.

Inverting mappings from smooth paths through Rn to paths through Rm: A technique applied to recovering articulation from acoustics

Motor theories, which postulate that speech perception is related to linguistically significant movements of the vocal tract, have guided speech perception research for nearly four decades but have had little impact on automatic speech recognition. In this paper, we describe a signal processing technique named MIMICRI that may help link motor theory with automatic speech recognition by providing a practical approach to recovering articulator positions from acoustics. MIMICRI’s name reflects three important operations it can perform on time-series data: it can reduce the dimensionality of a data set (manifold inference); it can blindly invert nonlinear functions applied to the data (mapping inversion); and it can use temporal context to estimate intermediate data (contextual recovery of information). In order for MIMICRI to work, the signals to be analyzed must be functions of unobservable signals that lie on a linear subspace of the set of all unobservable signals. For example, MIMICRI will typically work if the unobservable signals are band-pass and we know the pass-band, as is the case for articulator motions. We discuss the abilities of MIMICRI as they relate to speech processing applications, particularly as they relate to inverting the mapping from speech articulator positions to acoustics. We then present a mathematical proof that explains why MIMICRI can invert nonlinear functions, which it can do even in some cases in which the mapping from the unobservable variables to the observable variables is many-to-one. Finally, we show that MIMICRI is able to infer accurately the positions of the speech articulators from speech acoustics for vowels. Five parameters estimated by MIMICRI were more linearly related to articulator positions than 128 spectral energies.

A Model of Mandibular Movements During Speech: Normative Pilot Study for the Brazilian Portuguese Language

The precision of speech articulation is related to the possibility and freedom of the mandibular movements, modifying the spaces in order to allow the different articulatory positions of each sound. Electrognathography allows the objective delineation and registration of the mandibular movements, determining the level of opening, translations and velocity of these movements. Its use is a resource that can establish quantitative diagnostic parameters. The aim of this study was to verify the amplitude, velocity and characterization of the mandibular movements during speech using computerized electrognathography. Participants were 40 adults, male and female, with no temporomandibular disorders; with no missing teeth; with no dental occlusion alterations or dentofacial deformities; with no dental prostheses; and with no communication, neurological or cognitive deficits. The mandibular movements were observed during the sequential naming of pictures containing all the phonemes of the Brazilian Portuguese language. The registrations were obtained using electrognathography (BioENG – BioPak system), assessing the spatial position, course and velocity of the mandibular movements. The mean values of velocity were: 88.65 mm/sec during opening and 89.90mm/sec during closing. The mean values of amplitude were: sagittal opening: 12.77 mm, frontal opening: 11.21 mm, protrusion: 1.22 mm; retrusion 5.67 mm; translations to the right: 1.49 mm and to the left: 1.59 mm. The velocity of opening is directly related to that of closing. The amplitude of opening demonstrates a direct correlation with the velocity of opening and closing. All participants presented lateral translations during the course of the jaw. The assessment of speech in normal individuals is characterized by: discreet mandibular movements with an anteroposterior component and lateral translations. This study allowed for the delineation of a profile of the mandibular movements during speech in asymptomatic individuals.

Reliablity of measurements from ultrasound images

As ultrasound imaging gains popularity in phonetic and speech science research, examining the reliability of measures taken from ultrasound images becomes important. This study assesses the reliability of hand measures of ultrasound data collected by graduate student researchers at the University of South Florida ultrasound imaging lab. Speech production data from two different experiments, ‘‘Ultrasound analysis of velar fronting’’ (Wodzinski, 2004) and ‘‘Ultrasound study of errors in speech production’’ [Frisch, (2003)] were analyzed by two different researchers to obtain inter-rater reliability measures. In addition, one data set was measured twice by the same researcher, once when inexperienced with ultrasound analysis and 7 months later after considerable experience had been gained. The study compared researcher’s choice of image to analyze, the measures of the location of articulatory landmarks, and the measures used to quantify articulatory postures. Overall, hand measures of ultrasound images were found to be reliable. There were some differences in the absolute measures obtained, however, different researcher’s measures of the same data led to the same conclusions about articulation. In addition, it was found that the measurements of different researchers became more similar to one another with experience.

Blind inversion and the perception/production link

Whether humans recover articulator positions from acoustics in the course of speech perception has been debated for years. Some of the arguments against recovering articulator positions are relevant for machine speech recognition as well. However, techniques for blind inversion, in which the mapping from acoustics to articulation is learned without known articulator trajectories, were not considered when some of these arguments were being developed. We discuss the impact of blind inversion on the debate, and conclude that there are pragmatic reasons to recover articulation from acoustics, particularly when blind inversion is used and the mapping from articulation to acoustics is many-to-one.

Integration of articulatory and spectrum features based on the hybrid HMM/BN modeling framework

Most of the current state-of-the-art speech recognition systems are based on speech signal parametrizations that crudely model the behavior of the human auditory system. However, little or no use is usually made of the knowledge on the human speech production system. A data-driven statistical approach to incorporate this knowledge into ASR would require a substantial amount of data, which are not widely available since their acquisition is difficult and expensive. Furthermore, during recognition, it is nearly impossible to obtain observations of articulators movement. Thus, research on speech production mechanisms in ASR has largely focused on modeling the hidden articulatory trajectories and using prior phonetic and phonological knowledge. Nevertheless, it has been shown that combining the acoustic and articulatory information can lead to improved speech recognition performance. The approach taken in this study is to integrate features extracted from actual articulatory data with acoustic MFCC features in a way that allows recognition using MFCC only. Rather than trying to map articulatory features to the corresponding acoustic features, we use the probabilistic dependency between them. Bayesian Networks (BN) are ideally suited for this purpose. They can model complex joint probability distributions with many discrete and continuous variables and have great flexibility in representing their dependencies. Our speech recognition system is based on the hybrid HMM/BN acoustic model where the BN is used to describe the HMM states’ probability distributions. HMM transitions, on the other hand, model the temporal speech characteristics. Articulatory and acoustic features are represented by different variables of the BN. Dependencies are learned from the observable articulatory and acoustic training data. During recognition, when only the acoustic observations are available, articulatory variables are assumed hidden. We have evaluated our ASR system by using a small database consisting of articulatory and acoustic data recorded from three speakers. The articulatory data are actual measurements of articulators position at several points. In all experiments involving both speaker-dependent and multi-speaker acoustic models, the HMM/BN system outperformed the baseline HMM system trained on acoustic data only. In experimenting with different BN topologies, we found that integrating the velocity and acceleration coefficients calculated as first and second derivatives of the articulatory position data can further improve recognition performance.

Anatomy of the Cricothyroid Articulation: Differences between Men and Women

A greater difficulty in exposing the arytenoid cartilage during rotation surgeries has been observed for men. The objective of the present study was to describe the position of the cricothyroid articulation and the distance between the right and left articulations, and to compare these findings between genders. The following measurements were obtained for 16 cadavers (9 men and 7 women): angulation of the cricothyroid articulation in the cricoid ring, dimensions of the thyroid articular surface of the cricoid cartilage, and distance between the cricothyroid and cricoarytenoid articulations. The cricothyroid articulation angle was narrower in men than in women (p = .04). The major diameter of the articular facet of the thyroid cartilage was wider in men (p = .001). The longer lamina of the thyroid cartilage, as well as the more posterior position of the cricothyroid articulation, in men might explain the greater difficulty in exposing the arytenoid cartilage during laryngeal framework surgeries observed for this group of patients.

Language-Specific Articulatory Settings: Evidence from Inter-Utterance Rest Position

The possible existence of language-specific articulatory settings (underlying or default articulator positions) has long been discussed, but these have proven elusive to direct measurement. This paper presents two experiments using X-ray data of 5 English and 5 French subjects linking articulatory setting to speech rest position, which is measurable without segmental interference. Results of the first experiment show that speech rest position is significantly different across languages at 5 measurement locations in the vocal tract, and is similar to previously described language-specific articulatory settings. The second experiment shows that the accuracy of achievement of speech rest position is similar to that of a specified vowel target (/i/). These results have implications for the phonetics and phonology of neutral vowels, segmental nventories, and L2 acquisition.

A neural network model of the articulatory-acoustic forward mapping trained on recordings of articulatory parameters

Three neural network models were trained on the forward mapping from articulatory positions to acoustic outputs for a single speaker of the Edinburgh multi-channel articulatory speech database. The model parameters (i.e., connection weights) were learned via the backpropagation of error signals generated by the difference between acoustic outputs of the models, and their acoustic targets. Efficacy of the trained models was assessed by subjecting the models' acoustic outputs to speech intelligibility tests. The results of these tests showed that enough phonetic information was captured by the models to support rates of word identification as high as 84%, approaching an identification rate of 92% for the actual target stimuli. These forward models could serve as one component of a data-driven articulatory synthesizer. The models also provide the first step toward building a model of spoken word acquisition and phonological development trained on real speech.

An articulation model for audiovisual speech synthesis—Determination, adjustment, evaluation

The authors present a visual articulation model for speech synthesis and a method to obtain it from measured data. This visual articulation model is integrated into MASSY, the Modular Audiovisual Speech SYnthesizer, and used to control visible articulator movements described by six motion parameters: one for the up-down movement of the lower jaw, three for the lips and two for the tongue. The visual articulation model implements the dominance principle as suggested by Löfqvist [Löfqvist, A., 1990. Speech as audible gestures. In: Hardcastle, W.J., Marchal, A. (Eds.), Speech Production and Speech Modeling. Kluwer Academic Publishers, Dodrecht, pp. 289–322]. The parameter values for the model derive from measured articulator positions. To obtain these data, the articulation movements of a female speaker were measured with the 2D-articulograph AG100 and simultaneously filmed. The visual articulation model is adjusted and evaluated by testing word recognition in noise.

Segments and segmental properties in cross-language perception: Korean perception of English obstruents in various prosodic locations

Experimental models of cross-language perception and second-language acquisition (such as PAM and SLM) typically treat language differences in terms of whether the two languages share phonological segmental categories. Linguistic models, by contrast, generally examine properties which cross classify segments, such as features, rules, or prosodic constraints. Such models predict that perceptual patterns found for one segment will generalize to other segments of the same class. This paper presents perceptual identifications of Korean listeners to a set of voiced and voiceless English stops and fricatives in various prosodic locations to determine the extent to which such generality occurs. Results show some class-general effects; for example, voicing identification patterns generalize from stops, which occur in Korean, to nonsibilant fricatives, which are new to Korean listeners. However, when identification is poor, there are clear differences between segments within the same class. For example, in identifying stops and fricatives, both point of articulation and prosodic position bias perceptions; coronals are more often labeled fricatives, and syllable initial obstruents are more often labeled stops. These results suggest that class-general perceptual patterns are not a simple consequence of the structure of the perceptual system, but need to be acquired by factoring out within-class differences.

Multiple effects of consonant manner of articulation and intonation type on F0 in English

In this study we examine how consonant manner of articulation interacts with intonation type in shaping the F0 contours in English. Native speakers of American English read aloud words differing in vowel length, consonant manner of articulation and consonant position in word. They produced each word in either a statement or question carrier. F0 contours of their speech were extracted by measuring every complete vocal period. Preliminary results based on graphic analysis of three speakers’ data suggest that there are three distinct consonantal effects: F0 interruption due to devoicing, a large but brief (10–40 ms) F0 raising at the onset of voicing, and a smaller but longer-lasting F0 raising throughout a large proportion of the preceding and following vowels. These effects appear to be imposed on a continuously changing F0 curve that is either rising-falling or falling-rising, depending on whether the carrier sentence is a statement or a question. Further analysis will test the hypothesis that these continuous curves result from local pitch targets that are assigned to individual syllables and implemented with them in synchrony regardless of their segmental composition. [Work supported by NIDCD Grant No. R01 DC03902.]

Aeroacoustics production of fricative speech sounds

The aeroacoustic production of fricative speech sounds is studied using schlieren imaging and acoustic measurements. The focus is the structure of the turbulent jets formed during fricative speech sound production, and how interaction of the jet with articulators affects the acoustic nature of the speech sound. Patterns of the jets formed during the articulation of both voiced and unvoiced fricatives (s and z) are shown using schlieren images. In particular, the interaction of the jet with articulators such as teeth and lips are clearly seen, and demonstrated by varying articulator positions. Pressure measurements were made in conjunction with the images using a microphone placed near the teeth and one in the farfield. The pressure measurements show the acoustic consequences of the various jet/articulator interactions, further clarifying which articulators are most important in determining the aeroacoustic source characteristics.