Vocal Tract Shape Research Articles

Automatic generation of the complete vocal tract shape from the sequence of phonemes to be articulated

Articulatory speech synthesis requires generating realistic vocal tract shapes from the sequence of phonemes to be articulated. This work proposes the first model trained from rt-MRI films to automatically predict all of the vocal tract articulators’ contours. The data are the contours tracked in the rt-MRI database recorded for one speaker. Those contours were exploited to train an encoder–decoder network to map the sequence of phonemes and their durations to the exact gestures performed by the speaker. Different from other works, all the individual articulator contours are predicted separately, allowing the investigation of their interactions. We measure four tract variables closely coupled with critical articulators and observe their variations over time. The test demonstrates that our model can produce high-quality shapes of the complete vocal tract with a good correlation between the predicted and the target variables observed in rt-MRI films, even though the tract variables are not included in the optimization procedure.

Acoustic Change Over Time in Spastic and/or Flaccid Dysarthria in Motor Neuron Diseases

This study aims to investigate acoustic change over time as biomarkers to differentiate among spastic-flaccid dysarthria associated with amyotrophic lateral sclerosis (ALS), spastic dysarthria associated with primary lateral sclerosis (PLS), flaccid dysarthria associated with spinal and bulbar muscular atrophy (SBMA), and to explore how these acoustic parameters are affected by dysarthria severity. Thirty-three ALS patients with mixed flaccid-spastic dysarthria, 17 PLS patients with pure spastic dysarthria, 18 SBMA patients with pure flaccid dysarthria, and 70 controls, all French speakers, were included in the study. Speakers produced vowel-glide sequences targeting different vocal tract shape changes. The mean and coefficient of variation of the total squared change of mel frequency cepstral coefficients were used to capture the degree and variability of acoustic changes linked to vocal tract modifications over time. Differences in duration of acoustic events were also measured. All pathological groups showed significantly less acoustic change compared to controls, reflecting less acoustic contrast in sequences. Spastic and mixed spastic-flaccid dysarthric speakers showed smaller acoustic changes and slower sequence production compared to flaccid dysarthria. For dysarthria subtypes associated with a spastic component, reduced degree of acoustic change was also associated with dysarthria severity. The acoustic parameters partially differentiated among the dysarthria subtypes in relation to motor neuron diseases. While similar acoustic patterns were found in spastic-flaccid and spastic dysarthria, crucial differences were found between these two subtypes relating to variability. The acoustic patterns were much more variable in ALS. This method forms a promising clinical tool as a diagnostic marker of articulatory impairment, even at mild stage of dysarthria progression in all subtypes.

Voice source, formant frequencies and vocal tract shape in overtone singing. A case study

Purpose: In overtone singing a singer produces two pitches simultaneously, a low-pitched, continuous drone plus a melody played on the higher, flutelike and strongly enhanced overtones of the drone. The purpose of this study was to analyse underlying acoustical, phonatory and articulatory phenomena. Methods: The voice source was analyzed by inverse filtering the sound, the articulation from a dynamic MRI video of the vocal tract profile, and the lip opening from a frontal-view video recording. Vocal tract cross-distances were measured in the MR recording and converted to area functions, the formant frequencies of which computed. Results: Inverse filtering revealed that the overtone enhancement resulted from a close clustering of formants 2 and 3. The MRI material showed that for low enhanced overtone frequencies (F E) the tongue tip was raised and strongly retracted, while for high F E the tongue tip was less retracted but forming a longer constriction. Thus, the tongue configuration changed from an apical/anterior to a dorsal/posterior articulation. The formant frequencies derived from the area functions matched almost perfectly those used for the inverse filtering. Further, analyses of the area functions revealed that the second formant frequency was strongly dependent on the back cavity, and the third on the front cavity, which acted like a Helmholtz resonator, tuned by the tongue tip position and lip opening. Conclusions: This type of overtone singing can be fully explained by the well-established source-filter theory of voice production, as recently found by Bergevin et al. [1] for another type of overtone singing.

Discrete constriction locations describe a comprehensive range of vocal tract shapes in the Maeda model.

The Maeda model was used to generate a large set of vocoid-producing vocal tract configurations. The resulting dataset (a) produced a comprehensive range of formant frequencies and (b) displayed discrete tongue body constriction locations (palatal, velar/uvular, and lower pharyngeal). The discrete parameterization of constriction location across the vowel space suggests this is likely a fundamental characteristic of the human vocal tract, and not limited to any specific set of vowel contrasts. These findings suggest that in addition to established articulatory-acoustic constraints, fundamental biomechanical constraints of the vocal tract may also explain such discreteness.

Effect of cumulative distance in motor space on segment-sized slips in 3-word phrases

Slips, or speech errors (e.g., “teep a cape” for “keep a tape”), have been described as the disordered arrangement of discrete elements, giving rise to the hypothesis that these arise from misordering during phonological encoding. An alternative hypothesis is that slips are articulatory timing errors, conditioned by the same motor factors that underlie coarticulation. This hypothesis motivates the present study, which used a modified slip procedure to elicit errors from 40 native English-speaking adults. Participants spoke 3-word phrases that were constructed as mini tongue twisters, with alternating word onsets and repeating rhymes. These phrases were designed to condition slips to a greater or less extent according to coarticulatory-adjacent processes (e.g., palatalization). A metric of cumulative distance-traveled in motor space during phrase production was also calculated: the degree of change in vocal tract shapes across the phrase, estimated using data generated within the tubesounds model in ArtiSynth. Phrases designed to more strongly condition slips were expected to cover less cumulative distance in motor space than phrases designed to less strongly condition slips. The distance-traveled metric did not always align with this expectation, but was nonetheless a better predictor of presence/absence of slips than phrase condition in analyses on preliminary data.

Measurement of Pharyngeal Air Pressure During Phonation Using High-Resolution Manometry.

Purpose The study of air pressure in the vocal tract is essential to understanding vocal function. Changes in vocal tract shape during different phonatory gestures are hypothesized to produce nonuniform air pressure across lower vocal tract locations. Current methods of air pressure measurement, however, are limited to a single location in the anterior oral cavity. The purposes of this study were (a) to assess the feasibility of a novel method of simultaneously measuring phonatory air pressure at multiple locations across the lower vocal tract using high-resolution pharyngeal manometry (HRM) and (b) to compare pressure across locations and among phonatory tasks. Method Two subjects underwent HRM while performing phonatory tasks. A catheter was passed transnasally and air pressure was measured simultaneously at five locations between the velopharyngeal port and the upper esophageal sphincter. Descriptive statistics were calculated for each location by task, and for each task averaged across locations. Results HRM was well tolerated, and air pressures from multiple locations in the lower vocal tract were able to be obtained simultaneously. During vocal tract semi-occlusion tasks, air pressures differed by location. Pressures averaged across locations demonstrated a pattern of increasing pressure with increasing semi-occlusion. Conclusions HRM is feasible for measuring air pressure simultaneously at multiple locations in the lower vocal tract during phonation with high spatial and temporal resolution, providing rich data to augment understanding of vocal function. The high spatial and temporal resolution yielded by this new method, paired with preliminary evidence that pressures change by location as a function of phonatory task, may be useful in future assays exploring differences in lower vocal tract air pressures between normal and disordered populations.

Dark tone quality and vocal tract shaping in soprano song production: Insights from real-time MRI.

Tone quality termed “dark” is an aesthetically important property of Western classical voice performance and has been associated with lowered formant frequencies, lowered larynx, and widened pharynx. The present study uses real-time magnetic resonance imaging with synchronous audio recordings to investigate dark tone quality in four professionally trained sopranos with enhanced ecological validity and a relatively complete view of the vocal tract. Findings differ from traditional accounts, indicating that labial narrowing may be the primary driver of dark tone quality across performers, while many other aspects of vocal tract shaping are shown to differ significantly in a performer-specific way.

Complexity of vocal tract shaping in glossectomy patients and typical speakers: A principal component analysis.

The glossectomy procedure, involving surgical resection of cancerous lingual tissue, has long been observed to affect speech production. This study aims to quantitatively index and compare complexity of vocal tract shaping due to lingual movement in individuals who have undergone glossectomy and typical speakers using real-time magnetic resonance imaging data and Principal Component Analysis. The data reveal that (i) the type of glossectomy undergone largely predicts the patterns in vocal tract shaping observed, (ii) gross forward and backward motion of the tongue body accounts for more change in vocal tract shaping than do subtler movements of the tongue (e.g., tongue tip constrictions) in patient data, and (iii) fewer vocal tract shaping components are required to account for the patients' speech data than typical speech data, suggesting that the patient data at hand exhibit less complex vocal tract shaping in the midsagittal plane than do the data from the typical speakers observed.

Using browser-based software to teach acoustic theory of speech production

I will demonstrate how I use browser-based software authored by Mark Huckvale to promote active learning of the acoustic theory of speech production. Students create the complex glottal tone using ESynth, filter the glottal tone with various vocal tract filters with ESystem, explore the impact of vocal tract shapes on formants using VTDemo, and examine their own vowel formants with ESection.

Acoustically Induced Vocal Training for Individuals With Impaired Hearing

Articulation, phonation, and resonance disorders in the speech of hearing-impaired-speakers reduces intelligibility. The study focusses on (1) whether nonacoustic feedback may facilitate the adjustment of the vocal tract, leading to increased vocal tract resonance, and (2) whether training with the feedback would be helpful for the subsequent formation of vowels. Prospective. Seven profoundly hearing-impaired participants used acoustic sound waves in the frequency range of the first two vocal tract resonances applied in front of the open mouth at intensities above 1 Pa. They were asked to amplify the sound via adjusting the vocal tract. The sound waves corresponded to the first and second resonance frequencies of the vowels [u], [o], and [a]. The self-assessment of the participants and a software-based/auditory analysis was reported. The participants were able to enhance the acoustic signal by adjusting the vocal tract shape. The self-perception of the participants, the auditory voice analysis, and the acoustic analysis of vowels were consistent with each other. While the maximum sound pressure levels were constant, the mean sound pressure levels increased. Breathiness and hoarseness declined during the exercises. Resonance/harmonic-to-noise ratio increased, especially for the vowels [u], [o], [a]. Furthermore, the positively connoted feedback from the participants indicated easier sound production. Nonauditory feedback, based on acoustic waves, could be suitable for improving the formation of vowels. The findings are in accordance with a reduction of acoustic losses within the vocal tract.

An inverse problem to determine the shape of a human vocal tract

The inverse problem of determining the cross-sectional area of a human vocal tract during the utterance of a vowel is considered. The frequency-dependent boundary condition at the lips is expressed in terms of the acoustic impedance of a vibrating piston on an infinite plane baffle. The corresponding pressure at the lips is expressed in terms of the normalized impedance and a key quantity related to the so-called Jost function of an associated Schrödinger equation. Various input data sets are considered to solve the inverse problem, including the pressure at the lips, the absolute pressure at the lips, and the poles of the pressure at the lips. The solution to the inverse problem is obtained with the help of the Gel’fand–Levitan method for the associated Schrödinger equation.

Objective Measures of Two Musical Interpretations of an Excerpt From Berlioz's “La mort d'Ophélie”

This study aimed to determine objective production differences relative to two emotional interpretations in performing an excerpt from a classical art song. The null hypothesis was proposed. The first author recorded an excerpt from an art song. The excerpt was sung with two contrasting musical interpretations: an "empathetic legato" approach, and a "sarcastic" approach characterized by emphatic attacks. Microphone, airflow, and electroglottography signals were digitized. The vowels were analyzed in terms of intensity, long term average spectra, fundamental frequency (fo), airflow vibrato rate and extent, vowel onset slope, intensity comparison of harmonic frequencies, and glottal measures based on electroglottograph waveforms. Four consonant tokens were analyzed relative to airflow, voice onset time, and production duration. The emphatic performance had faster vowel onset, increased glottal adduction, increased intensity of harmonics in 2-3 kHz, increased intensity in the fourth and fifth formants, inferred subglottal pressure increase, increased airflow for /f/, and greater aspiration airflow for /p, t/. Vibrato extents for intensity, fo, and airflow were wider in the emphatic approach. Findings revealed larger EGGW25 and peak-to-peak amplitude values of the electroglottography waveform, suggesting greater vocal fold contact area and longer glottal closure for the emphatic approach. Long-term average spectrum analyses of the entire production displayed minor variation across all formant frequencies, suggesting an insignificant change in vocal tract shaping between the two approaches. This single-case objective study emphasizes the reality of physiological, aerodynamic, and acoustic production differences in the interpretive and pedagogical aspects of art song performance.

$F_0$-Noise-Robust Glottal Source and Vocal Tract Analysis Based on ARX-LF Model

This paper proposes a robust automatic speech analysis method based on a source-filter model constructed of an Auto-Regressive eXogenous (ARX) model and the Liljencrants-Fant (LF) model. The proposed method estimates glottal source waveform and vocal tract shape parameters using an analysis-by-synthesis approach. Structurally, the first step is to initialize the glottal source parameters using the inverse filter method, and the second step is to simultaneously estimate the glottal source waveform and the vocal tract shape parameters using an analysis-by-synthesis approach with an iterative algorithm. The proposed method was verified on synthetic voices with different glottal noise (signal to noise ratio) from 0 dB to 50 dB and different fundamental frequency ( $F_0$ ) from 80 Hz to 320 Hz levels. The results show that the proposed method achieved a much higher estimation accuracy than that of the state-of-the-art inverse filtering methods on both different glottal noise and different $F_0$ levels.

Variation of Formant Frequency with Nepalese Vowels

The regions of speech spectrum in which the frequency corresponds to relatively large amplitude are known as formants. For any vocalic sounds, number of formants may occur in the frequency range 0 to 4000 Hz. The formant frequencies of speech sounds are directly depending up on the shape and size of vocal tract. The aim of study was to study the variation of formant frequency with Nepalese vowels. Ten Nepalese vowels word in initial position /VC/ as spoken three times by 10 male and 10 female Nepali speakers were recorded in system in the free field of partially acoustically treated room. PRRAT software is used to digitize and analyze the data. Linear predictive coding (LPC) spectra were obtained for each of vowels and formant frequencies were measured. By plotting curve between formant frequencies and vowels, explain their variation.

Using visual feedback to tune the second vocal tract resonance for singing in the high soprano range

Purpose Over a range roughly C5–C6, sopranos usually tune their first vocal tract resonance (R1) to the fundamental frequency (f o) of the note sung: R1:f o tuning. Those who sing well above C6 usually adjust their second vocal tract resonance (R2) and use R2:f o tuning. This study investigated these questions: Can singers quickly learn R2:f o tuning when given suitable feedback? Can they subsequently use this tuning without feedback? And finally, if so, does this assist their singing in the high range? Methods New computer software for the technique of resonance estimation by broadband excitation at the lips was used to provide real-time visual feedback on f o and vocal tract resonances. Eight sopranos participated. In a one-hour session, they practised adjusting R2 whilst miming (i.e. without phonating), and then during singing. Results Six sopranos learned to tune R2 over a range of several semi-tones, when feedback was present. This achievement did not immediately extend their singing range. When the feedback was removed, two sopranos spontaneously used R2:f o tuning at the top of their range above C6. Conclusions With only one hour of training, singers can learn to adjust their vocal tract shape for R2:f o tuning when provided with visual feedback. One additional participant who spent considerable time with the software, acquired greater skill at R2:f o tuning and was able to extend her singing range. A simple version of the hardware used can be assembled using basic equipment and the software is available online.

Acoustic and articulatory analysis and synthesis of shouted vowels

Acoustic and articulatory differences between spoken and shouted vowels were analyzed for two male and two female subjects by means of acoustic recordings and midsagittal magnetic resonance images of the vocal tract. In accordance with previous acoustic findings, the fundamental frequencies, intensities, and formant frequencies were all generally higher for shouted than for spoken vowels. The harmonics-to-noise ratios and H1-H2 measures were generally lower for shouted vowels than for spoken vowels. With regard to articulation, all subjects used an increased lip opening, an increased jaw opening, and a lower tongue position for shouted vowels. However, the changes of vertical larynx position, uvula elevation, and jaw protrusion between spoken and shouted vowels were inconsistent among subjects. Based on the analysis results, a perception experiment was conducted to examine how changes of fundamental frequency, subglottal pressure, vocal tract shape, and phonation type contribute to the perception of stimuli created by articulatory synthesis as being shouted. Here, fundamental frequency had the greatest effect, followed by vocal tract shape and lung pressure, with no measurable effect of phonation type.

Articulation and identification of voiced stop consonants produced by acoustically driven vocal tract modulations

A recently developed model of speech production [Story & Bunton, JASA, 146(4), 2522–2528] was used to generate VCVs that were examined with regard to both articulation and identification of the consonant. In this model, an utterance is generated by specifying relative acoustic events along a time axis. These events consist of directional changes of the vocal tract resonance frequencies called resonance deflection patterns (RDPs) that, when associated with a temporal event function, are transformed via acoustic sensitivity functions, into time-varying modulations of the vocal tract shape. RDPs specifying /b/, /d/, and /g/ would typically be coded as [−1 −1 −1], [−1 1 1], and [−1 1 −1], respectively, indicating, from left to right, the targeted directional shift of the first, second, and third resonances of the vocal tract. In this study, two types of V1CV2 continua were constructed in three vowel contexts (/i, a, u/) by incrementing in small steps (1) the second resonance deflection from −1 to 1, and (2) the third resonance deflection from 1 to −1. The resulting time-varying vocal tract shapes emulate expected articulation patterns for the stop consonants, and a perceptual experiment indicated that listeners identify the consonants based on the polarity of RDP values.

Printable 3D vocal tract shapes from MRI data and their acoustic and aerodynamic properties

A detailed understanding of how the acoustic patterns of speech sounds are generated by the complex 3D shapes of the vocal tract is a major goal in speech research. The Dresden Vocal Tract Dataset (DVTD) presented here contains geometric and (aero)acoustic data of the vocal tract of 22 German speech sounds (16 vowels, 5 fricatives, 1 lateral), each from one male and one female speaker. The data include the 3D Magnetic Resonance Imaging data of the vocal tracts, the corresponding 3D-printable and finite-element models, and their simulated and measured acoustic and aerodynamic properties. The dataset was evaluated in terms of the plausibility and the similarity of the resonance frequencies determined by the acoustic simulations and measurements, and in terms of the human identification rate of the vowels and fricatives synthesized by the artificially excited 3D-printed vocal tract models. According to both the acoustic and perceptual metrics, most models are accurate representations of the intended speech sounds and can be readily used for research and education.

SongF0: A Spectrum-Based Fundamental Frequency Estimation for Monophonic Songs

In this paper, we propose a fundamental frequency ( $$f_0$$ ) estimation method for monophonic songs. The proposed method (SongF0) performs simultaneous voiced/unvoiced detection (VUD) and $$f_0$$ estimation from the frequency spectrum. Even though the spectrogram exhibits ample information of the singer $$f_0$$ in the form of harmonic partials, the existing spectrum-based $$f_0$$ detection methods fail to accurately extract the $$f_0$$ due to the complex singing styles. In order to cover a wider range of fundamental frequencies, singers adjust their vocal tract shape by lowering the larynx and widening the pharynx results in tuning lower harmonic partial to formant frequencies (Sundberg in STL-QPSR 1:1–6, 1968; Speech Transm Lab Q Prog Status Rep 4:21–39, 1970). Hence, most of the available popular $$f_0$$ detection methods which are inspired by speech production mechanism are susceptible to the energized higher-order harmonic spectral partials near the formants. In this work, we profoundly explore the quasi-harmonic nature of the spectral peaks to minimize the effect of formant frequencies on the detected $$f_0$$ . Initially, we train an ensemble classifier with the novel spectral features to predict the candidate $$f_0$$ harmonic partials from the spectrum. We exploit the property of constant spectral distance between the harmonic partials to reliably extract the singing $$f_0$$ from the predicted harmonic partials. We propose novel post-processing methods to significantly improve the $$f_0$$ detection accuracy in the weakly voiced and inharmonic transition regions. The proposed SongF0 which is independent of the vocal $$f_0$$ range is compared with the state-of-the-art $$f_0$$ extraction methods proposed for both speech and singing voice. The evaluation results on the openly available singing $$f_0$$ gold standard datasets revealed that the proposed method is significantly better than the several state-of-the-art $$f_0$$ detection methods.

How an aglossic speaker produces an alveolar-like percept without a functional tongue tip.

It has been previously observed [McMicken, Salles, Berg, Vento-Wilson, Rogers, Toutios, and Narayanan. (2017). J. Commun. Disorders, Deaf Stud. Hear. Aids 5(2), 1-6] using real-time magnetic resonance imaging that a speaker with severe congenital tongue hypoplasia (aglossia) had developed a compensatory articulatory strategy where she, in the absence of a functional tongue tip, produced a plosive consonant perceptually similar to /d/ using a bilabial constriction. The present paper provides an updated account of this strategy. It is suggested that the previously observed compensatory bilabial closing that occurs during this speaker's /d/ production is consistent with vocal tract shaping resulting from hyoid raising created with mylohyoid action, which may also be involved in typical /d/ production. Simulating this strategy in a dynamic articulatory synthesis experiment leads to the generation of /d/-like formant transitions.