Vocal Tract Area Research Articles

A New Method for Obtaining Accurate Estimates of Vocal-Tract Filters and Glottal Waves From Vowel Sounds

Previously, estimating vocal-tract filters and glottal waves from vowel sounds imposed either the invalid assumption that glottal waves over closed glottal intervals are zero, or parametric models for glottal waves, resulting in biased vocal-tract-filter estimates and glottal-wave estimates lacking information over closed glottal intervals. We obtain unbiased vocal-tract-filter estimates from sustained vowel sounds, for which the glottal waveforms are periodically stationary random processes. Two equations are constructed each relating the vocal-tract filter to the sound signal and the glottal wave over one of two closed glottal intervals. By subtracting one equation from the other, the periodic components of the glottal wave are eliminated from the vocal-tract-filter estimation, and an unbiased vocal-tract-filter estimate is obtained. The average of many such estimates from different closed glottal intervals of the sound is the final estimate, which is used to obtain the glottal wave by inverse filtering the sound. The results obtained from vowel sounds /a/ produced by some subjects are presented. Over closed glottal phases, the glottal waves obtained are nonzero. During vocal-fold colliding, they increase; during vocal-fold parting, they decrease or even increase. The vocal-tract filters obtained yield vocal-tract area functions similar to that measured from an unknown subject's magnetic resonance image.

Technique for “tuning” vocal tract area functions based on acoustic sensitivity functions

A technique for modifying vocal tract area functions is developed by using sum and difference combinations of acoustic sensitivity functions to perturb an initial vocal tract configuration. First, sensitivity functions [e.g., Fant and Pauli, Proc. Speech Comm. Sem. 74, 1975] are calculated for a given area function, at its specific formant frequencies. The sensitivity functions are then multiplied by scaling coefficients that are determined from the difference between a desired set of formant frequencies and those supported by the current area function. The scaled sensitivity functions are then summed together to generate a perturbation of the area function. This produces a new area function whose associated formant frequencies are closer to the desired values than the previous one. This process is repeated iteratively until the coefficients are equal to zero or are below a threshold value.

Measurement of temporal changes in vocal tract area function from 3D cine-MRI data

A 3D cine-MRI technique was developed based on a synchronized sampling method [Masaki et al., J. Acoust. Soc. Jpn. E 20, 375-379 (1999)] to measure the temporal changes in the vocal tract area function during a short utterance /aiueo/ in Japanese. A time series of head-neck volumes was obtained after 640 repetitions of the utterance produced by a male speaker, from which area functions were extracted frame-by-frame. A region-based analysis showed that the volumes of the front and back cavities tend to change reciprocally and that the areas near the larynx and posterior edge of the hard palate were almost constant throughout the utterance. The lower four formants were calculated from all the area functions and compared with those of natural speech sounds. The mean absolute percent error between calculated and measured formants among all the frames was 4.5%. The comparison of vocal tract shapes for the five vowels with those from the static MRI method suggested a problem of MRI observation of the vocal tract: data from static MRI tend to result in a deviation from natural vocal tract geometry because of the gravity effect.

Synergistic modes of vocal tract articulation for American English vowels

The purpose of this study was to investigate the spatial similarity of vocal tract shaping patterns across speakers and the similarity of their acoustic effects. Vocal tract area functions for 11 American English vowels were obtained from six speakers, three female and three male, using magnetic resonance imaging (MRI). Each speaker's set of area functions was then decomposed into mean area vectors and representative modes (eigenvectors) using principal components analysis (PCA). Three modes accounted for more than 90% of the variance in the original data sets for each speaker. The general shapes of the first two modes were found to be highly correlated across all six speakers. To demonstrate the acoustic effects of each mode, both in isolation and combined, a mapping between the mode scaling coefficients and [F1, F2] pairs was generated for each speaker. The mappings were unique for all six speakers in terms of the exact shape of the [F1, F2] vowel space, but the general effect of the modes was the same in each case. The results support the idea that the modes provide a common system for perturbing a unique underlying neutral vocal tract shape.

Common modes of vocal tract articulation for vowels

The purpose of this study was to investigate the spatial similarity of vocal tract shaping patterns across speakers and the similarity of their acoustic effects. Vocal tract area functions for eleven American English vowels were first obtained from six speakers, three female and three male, using magnetic resonance imaging (MRI). Each speaker’s set of area functions was then decomposed into mean area vectors and representative modes (eigenvectors) using principal components analysis (PCA). Three modes accounted for more than 90% of the variance in the original data sets for each speaker. The general shapes of the first two modes were found to be highly correlated across all six speakers. To demonstrate the acoustic effects the modes, both isolated and combined, a mapping between the mode scaling coefficients and [F1,F2] pairs was generated for each speaker. The mappings were unique for all six speakers in terms of the exact shape of the [F1,F2] vowel space, but the general effect of the modes was the same in each case. The results tend to support the idea that the modes provide a common system for perturbing a unique underlying neutral vocal tract shape. [Work supported by NIH R01-DC04789.]

Vocal Tract Area Functions to Assist in Determining Voiced Epochs in Esophageal Speech

Esophageal talkers may have reduced intelligibility due to both time domain and frequency domain variability. The unpredictable nature of esophageal speech can cause problems when automatic procedures are used in applications such as long-distance telephone messages. The current study compared a standard coding algorithm (LPC-10e) with a novel approach to determining voiced periods (vocal tract area functions) in the speech of esophageal talkers. The results of the study showed that the sentences synthesized with the vocal tract area function algorithm were more intelligible than those synthesized with the standard LPC-10e algorithm. Supplemental information, such as vocal tract area functions, may be useful in determining voiced epochs when variability in vocal parameters is high.

Une nouvelle méthode de mesure de la fonction d'aire du conduit vocal : cas des voyelles

Knowledge of vocal tract area functions is important for the understanding of phenomena occurring during speech production. We present here a new measurement method based on the external excitation of the vocal tract with a known pseudo-random sequence, where the area function is obtained by a linear prediction analysis applied to the cross-correlation between the sequence and the signal measured at the lips. The advantages of this method over methods based on sweep-tones or white noise excitation are (1) a much shorter measurement time (about 100 ms) and (2) the possibility of speech sound production during the measurement. This method has been checked against classical methods through systematic comparisons on a small corpus of vowels. Moreover, it has been verified that simultaneous speech sound production does not perturb significantly the measurements. This method should thus be a very helpful tool for the investigation of the acoustic properties of the vocal tract in various cases for vowels.

A parametric model of the vocal tract area function for vowel and consonant simulation

A model of the vocal-tract area function is described that consists of four tiers. The first tier is a vowel substrate defined by a system of spatial eigenmodes and a neutral area function determined from MRI-based vocal-tract data. The input parameters to the first tier are coefficient values that, when multiplied by the appropriate eigenmode and added to the neutral area function, construct a desired vowel. The second tier consists of a consonant shaping function defined along the length of the vocal tract that can be used to modify the vowel substrate such that a constriction is formed. Input parameters consist of the location, area, and range of the constriction. Location and area roughly correspond to the standard phonetic specifications of place and degree of constriction, whereas the range defines the amount of vocal-tract length over which the constriction will influence the tract shape. The third tier allows length modifications for articulatory maneuvers such as lip rounding/spreading and larynx lowering/raising. Finally, the fourth tier provides control of the level of acoustic coupling of the vocal tract to the nasal tract. All parameters can be specified either as static or time varying, which allows for multiple levels of coarticulation or coproduction.

Individual variation of the hypopharyngeal cavities and its acoustic effects

Morphological measurements of the hypopharynx are conducted to investigate the correlation between fine structures of the vocal tract and speaker characteristics. The hypopharynx includes the laryngeal tube and bilateral cavities of the piriform fossa. MRI data during sustained phonation of the five Japanese vowels by four subjects are obtained to analyze intra- and inter-speaker variation of the hypopharynx. Morphological analysis on the mid-sagittal and transverse planes revealed that the shape of the hypopharynx was relatively stable, regardless of vowel type, in contrast to relatively large inter-speaker variation, and these results are confirmed quantitatively by a simple similarity method. The small intra-speaker variation of the hypopharynx is confirmed by further morphological analysis using high-quality MRI data for one of the subjects, obtained by using the “phonation-synchronized method” and “custom laryngeal coil.” Furthermore, acoustical effects of the individual variation of the hypopharynx are estimated by using a transmission line model. Vocal tract area function of one of the subjects above the hypopharynx is combined with the hypopharyngeal cavities of other subjects, and their transfer functions are calculated. The results show that the inter-speaker variation of the hypopharynx affects spectra in the frequency range beyond approximately 2.5 kHz.

WinSingad: A real-time display for the singing studio

This paper describes the nature and implementation of a specially-designed integrated real-time display that is undergoing evaluation as part of a recently funded innovative pilot project to investigate the relative usefulness of computer displays in the singing studio. Following previous work that suggests that simple displays of a small number of analysis parameters are generally likely to be the most effective, the system makes available a range of complementary analyses that are plotted against time. These relate to: fundamental frequency, spectrum, spectral ratio, and vocal tract area. These can be viewed singly, multiply or in combination using a panel based design within the PC Windows environment, known as WinSingad. The algorithms used are described and the displays themselves are illustrated with results gained from the pilot phase of the research to indicate their potential usefulness.

A formant-to-area conversion technique based on acoustic sensitivity functions

A technique for transforming formant frequencies to vocal tract area functions is developed by using sum and difference combinations of acoustic sensitivity functions to perturb an initial vocal tract configuration. First, sensitivity functions [e.g., Fant and Pauli, Proc. Speech Comm. Sem., Vol. 74 (1975)] are calculated for a neutral area function, at its specific formant frequencies. The sensitivity functions are then multiplied by scaling coefficients that are determined from the difference between a desired set of formant frequencies and those supported by the current area function. Then, the scaled sensitivity functions are summed together to generate a perturbation of the area function. This produces a new area function whose associated formant frequencies are closer to the desired values than the previous one. This process is repeated iteratively until the coefficients are equal to zero or are below a threshold value. Results show that vowel area functions are generated that are comparable to those measured with imaging techniques. The performance of the technique with time-varying formant frequencies will also be discussed. [Work supported by NIH R01-DC04789.]

Gestural stability in vowels

In accordance with proper perception of linguistic sound units, past research has demonstrated some degree of acoustic and physiological stability. In contrast, articulatory stability has been thought to be inconsistent because articulations may vary so long as the vocal tract area function results in appropriate formant structure [Atal et al., J. Acoust. Soc. Am. 63, 1535–1555 (1978)]. However, if the area function for the constriction and its anterior region can maintain acoustic stability, articulatory stability should be observed in the relational behavior of four tongue pellets used in xray microbeam data. Previous work examined normalized pellet data in order to arrive at an average posture for each vowel [Hashi et al., J. Acoust. Soc. Am. 104, 2426–2437 (1998)]. But by assuming static (average) gestures, the research fell short of a correct postural characterization. This study of tongue pellet speed and normalized pellet displacement of front vowels spoken by ten microbeam database subjects reports that the tongue tip pellet speed maxima identify vowel edges (end of vowel onset, beginning of offset) while displacement of the three anterior pellets identify changes in formant structure (e.g., two stable regions in the Northern Cities English front low vowel).

A method of tooth superimposition on MRI data for accurate measurement of vocal tract shape and dimensions

Measurement of vocal tract area functions from MRI data requires a technique for tooth visualization because the teeth are as transparent as air in the images. In this article, a new method is proposed to accurately superimpose the teeth onto MRI volume data. Upper and lower tooth images with the surrounding bony structure are obtained by scanning a subject holding a contrast medium in the oral cavity. They are superimposed onto the target volume data via a three-dimensional transformation using landmarks sampled from the tooth images and target MRI data. The accuracy of the dental image superimposition is ensured by the minimization of the error volume, which is the mismatch volume of the dental image overlapping the surrounding soft tissue. The method is evaluated using five operators sampling the landmarks. Results show that the error volume is significantly reduced to a nearly constant value regardless of the operator’s skill.

Source and filter adjustments affecting the perception of the vocal qualities twang and yawn

Two vocal qualities, twang and yawn, were synthesized and rated perceptually. The stimuli consisted of synthesized vocal productions of a sentence-length utterance ‘ya ya ya ya ya,’ which had speech-like intonation. In a continuum transformation from normal to twang, the area in the pharynx was gradually decreased, along with vocal tract shortening and a decreased open quotient in the glottal airflow. In a continuum transformation toward yawn, the area in the pharynx was gradually increased, along with vocal tract lengthening and an increased open quotient. The normal (untransformed) vocal tract area was pre-determined by earlier studies involving MRI scans of a human subject's vocal tract. Listeners were asked to rate (on a scale from 1–10) the ‘amount of twang’ in one listening session and the ‘amount of yawn’ in another listening session. Overall, the perception of twang increased directly with pharyngeal area narrowing, vocal tract shortening, and decreased open quotient. The perception of yawn increased with pharyngeal area widening, vocal tract lengthening, and increased open quotient. Adjustments of one parameter alone yielded less significant perceptual changes than the above combinations, with open quotient showing the greatest effect in isolation. Listeners demonstrated variable perceptions in both continua with poor inter-subject, intra-subject, and inter-group reliability.

Simulation of VCV syllables with a parametric area function model of the vocal tract

A model of the vocal tract area function has been developed that consists of three tiers. The first tier is a vowel substrate defined by a system of spatial eigenmodes determined from MRI and/or x-ray microbeam data and a neutral area function. The input parameters to this first tier are coefficient values that, when multiplied by the appropriate eigenmode and added to the neutral area function, construct a desired vowel. The second tier consists of a Gaussian shaping function defined along the length of the vocal tract that modifies the vowel substrate such that a constriction is formed. Input parameters consist of the location, area, and range of the constriction. Location and area roughly correspond to the standard phonetic notions of place and degree of constriction while range defines the amount of vocal tract length over which the constriction will influence the tract shape. The third tier allows length modifications for articulatory maneuvers such as lip rounding/spreading and larynx lowering/raising. All parameters can be specified either as static or time-varying, which allows for multiple levels of coarticulation to be produced. The model will be discussed and then used to simulate several VCV syllables. [Work supported by NIH R01-DC04789.]

Simulation of singing qualities governed by lower vocal tract adjustments

In previous meetings, voice qualities such as pressed, ring, yawn, and twang were discussed in a speech context. It was shown that these qualities have unique spectral characteristics brought about by combinations of glottal and lower vocal tract adjustments (the epilarynx tube and the pharynx). Yawn has a wide glottis, a wide epilarynx tube, and a wide pharynx. On the contrary, twang has a general narrowing of all these airway sections. Ring has a wide pharynx and a relatively narrow epilarynx tube. A pressed voice is primary laryngeal, with a narrowed glottis. In this presentation, similar adjustments are made for singing with a voice simulator that controls vocal tract area functions and glottal flow pulses by rules. Results suggest that various singing styles, such as country‐western, opera, or pop, may in part be characterized by these unique combinations of source and filter adjustments.

Model‐based speaker normalization methods for speech recognition

AbstractA speaker normalization method using a speech generation model is proposed in order to achieve high‐performance speaker adaptation with a small amount of adaptation data. The speaker‐ and phoneme‐dependent vocal tract area function is approximated by the corresponding area function produced by the articulatory model of a standard speaker, combined with phoneme‐independent feature quantities of the vocal‐tract shape of the normalized target speaker as estimated from the formant frequencies of two vowels. The frequency warping functions are determined from the formant frequencies of speech calculated from the vocal‐tract area functions thus obtained, and normalization of the uttered speech is performed by stretching the speech spectrum in the frequency‐axis direction. Continuous phoneme recognition experiments using phoneme connection rules show that the recognition error using a gender‐dependent model is reduced by about 30% in the proposed method and that recognition performance superior to that of vocal‐tract length normalization is obtained. The recognition performance of the proposed method is also equivalent to that of speaker adaptation by moving vector field smoothing (VFS) using 10 phonetically balanced sentences, showing that high‐performance speaker adaptation using a small amount of adaptation data can be achieved by the proposed method. © 2003 Wiley Periodicals, Inc. Electron Comm Jpn Pt 2, 86(2): 45–56, 2003; Published online in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/ecjb.10119

A parametric area function model of three female vocal tracts based on orthogonal modes

Image sets corresponding to 11 vowels were obtained from each of 3 female speakers using Magnetic Resonance Imaging. From these images, vocal tract area functions for each vowel were determined. Each speaker’s 11 area functions were then decomposed into a set of orthogonal basis functions or ‘‘modes’’ and mean vowel area functions. Similar to previous work with male vocal tract data [Story and Titze, J. Phon. (1998)], two orthogonal modes accounted for nearly 90% of the variance within each speaker’s vowel set, while the mean area functions serve as neutral, but unique, tract shapes around which the vowels are formed. For each speaker, the two most significant modes and the mean area function were used to develop a parametric model that allowed for calculation of an arbitrary area function based on coefficients of the modes. With these models a hypothetical vowel space was generated in terms of F1, F2 pairs (first and second formant frequencies). The vowel spaces for each speaker were then used to map time-varying formant frequencies extracted from natural speech to the modal coefficients of their respective parametric models. This effectively produces a time-varying area function suitable for synthesis of speech. [Work supported by NIH R01-DC04789.]

A preliminary study of voice quality transformation based on modifications to the neutral vocal tract area function

The idea is pursued that voice quality can be partially represented by the underlying shape of a speaker's neutral vocal tract. Using an area function model, which allows direct access to the neutral tract shape, four separate modifications were made to one male speaker's vocal tract. The modifications involve imposing constrictive or expansive effects on the pharyngeal and oral portions of the neutral area function as well as on lip aperture and the epi-laryngeal tube. A single word utterance was first synthesized by superimposing deformation patterns appropriate for the word onto the original neutral tract shape (area function). Then, four additional samples of the word were synthesized using different modified neutral area function each time. The modifications were assessed by comparing F 1– F 2 formant trajectories of the original utterance with those of the modifications. The formant frequencies were observed to shift within the F 1– F 2 plane in directions predictable from simple tube acoustics. However, the modified voice qualities did not preserve the shape of the original F 1– F 2 trajectory. In other words, the modifications did not create a simple linear transformation of formant frequencies even though the “articulatory dynamics” (deformation patterns of the area function) were identical in all cases. These somewhat artificial vocal tract modifications were also compared with formant frequencies extracted from recordings of a speaker attempting to produce the same types of modifications. In general, the speaker's formant trajectories showed some similarities to the synthesized versions. However, the speaker also seemed to grade the “level” of the voice quality that was exerted on the utterance depending on whether the demands of the voice quality were in competition with the linguistic demands of a given phonetic segment. Finally, to demonstrate this type of voice quality modification in a broader context, the same procedures were applied to sentence-level speech and results were again shown as F 1– F 2 formant trajectories.

4D-MRI using the synchronized sampling method (SSM)

A synchronized sampling method (SSM) was developed for the study of voluntary movements by combining the electrocardiographic (ECG) gating method with an external triggering device, and four-dimensional magnetic resonance imaging (4D-MRI) at a rate of 30 frames per second was accomplished by volumetric imaging with the SSM. This method was first applied to the motion imaging of articulatory organs during repetitions of a Japanese five-vowel sequence, and the dynamic change in vocal tract area function was demonstrated with sufficient temporal resolution. This paper describes the methodology, applicability, and limitations of 4D-MRI with the SSM.