Vocal Tract Area Research Articles

Towards Improved Auditory-Perceptual Assessment of Timbres: Comparing Accuracy and Reliability of Four Deconstructed Timbre Assessment Models

Timbre is a central quality of singing, yet remains a complex notion poorly understood in psychoacoustic studies. Previous studies note how no single acoustic variable or combinations of variables consistently predict timbre dimensions. Timbre varies on a continuum from darkest to lightest. These extremes are associated with laryngeal and vocal tract adjustments related to smaller and larger vocal tract area and variations in vocal fold vibratory characteristics. Perceptually, timbre assessment is influenced by spectral characteristics and formant frequency adjustments, though these dimensions are not independently perceived. Perceptual studies repeatedly demonstrate difficulties in correlating variations in timbre stimuli to specific measures. A recent study demonstrated how acoustic predictive salience of voice category and voice weight across pitches contribute to timbre assessments and concludes that timbre may be related to as-of-yet unknown factor(s). The purpose of this study was to test four different models for assessing timbre; one model focused on specific anatomy, one on listener intuition, one utilising auditory anchors, and one using expert raters in a deconstructed timbre model with 5 specific dimensions. MethodsFour independent panels were conducted with separate cohorts of professional singing teachers. 41 assessors took part in the anatomically focused panel, 54 in the intuition-based panel, 30 in the anchored panel, and 12 in the expert listener panel. Stimuli taken from live performances of well-known singers were used for all panels, representing all genders, genres, and styles across a large pitch range. All stimuli are available as supplementary materials. Fleiss’ kappa values, descriptive statistics, and significance tests are reported for all panel assessments. ResultsPanel 1 through 4 varied in overall accuracy and agreement. The intuitionbased model showed overall 45% average accuracy (SD ±4%), k=0.289 (<0.001) compared to overall 71% average accuracy (SD ±3%), k=0.368 (<0,001) of the anatomical focused panel. The auditory-anchored model showed overall 75% average accuracy (SD ±8%), k=0.54 (<0.001) compared with overall 83% average accuracy and agreement of k=0.63 (<0.001) for panel 4. Results revealed that highest accuracy and reliability was achieved in a deconstructed timbre model and that providing anchoring improved reliability but with no further increase in accuracy. ConclusionDeconstructing timbre into specific parameters improved auditory perceptual accuracy and overall agreement. Assessing timbre along with other perceptual dimensions improves accruacy and reliability. Panel assessors’ expert level of listening skills remain an important factor in obtaining reliable and accurate assessments of auditory stimuli for timbre dimensions. Anchoring improved reliability but with no further increase in accuracy. The study suggests that timbre assessment can be improved by approaching the percept through a prism of 5 specific dimensions each related to specific physiology and auditory-perceptual subcategories. Further tests are needed with framework-naïve listeners, non-musically educated listeners, artificial intelligence comparisons, and synthetic stimuli to further test the reliability.

Sketches of chimpanzee (Pan troglodytes) hoo’s: vowels by any other name?

In human speech, the close back rounded vowel /u/ (the vowel in “boot”) is articulated with the tongue arched toward the dorsal boundary of the hard palate, with the pharyngeal cavity open. Acoustic and perceptual properties of chimpanzee (Pan troglodytes) hoo’s are similar to those of the human vowel /u/. However, the vocal tract morphology of chimpanzees likely limits their phonetic capabilities, so that it is unlikely, or even impossible, that their articulation is comparable to that of a human. To determine how qualities of the vowel /u/ may be achieved given the chimpanzee vocal tract, we calculated transfer functions of the vocal tract area for tube models of vocal tract configurations in which vocal tract length, length and area of a laryngeal air sac simulacrum, length of lip protrusion, and area of lip opening were systematically varied. The method described is principally acoustic; we make no claim as to the actual shape of the chimpanzee vocal tract during call production. Nonetheless, we demonstrate that it may be possible to achieve the acoustic and perceptual qualities of back vowels without a reconfigured human vocal tract. The results, while tentative, suggest that the production of hoo’s by chimpanzees, while achieving comparable vowel-like qualities to the human /u/, may involve articulatory gestures that are beyond the range of the human articulators. The purpose of this study was to (1) stimulate further simulation research on great ape articulation, and (2) show that apparently vowel-like phenomena in nature are not necessarily indicative of evolutionary continuity per se.

Bora's high vowels involve a two-way dental contrast, not a three-way backness contrast

The Bora language of Peru has six phonemic vowels, conventionally transcribed as /i ε a o ɨ ɯ/. This inventory is noteworthy since it appears to exhibit a three-way backness distinction among three high unrounded vowels. In this paper we present audiovisual data which confirms two significant facts: (1) /ɯ/ is high, back, and unrounded; and (2) the only vowel produced with lip rounding is /o/. However, the experiment further reveals that the Bora segment traditionally written /ɨ/ is actually not central but rather a possibly novel type of front vowel, articulated with substantial contact between the tongue and both the upper and lower teeth. We use the methodology of Iskarous (2010) to estimate vocal tract area functions for Bora vowels based on their observed formant frequencies and amplitudes. The results of this procedure support our claim that /ɨ/ is front and /ɯ/ is back. Several phonological alternations also motivate this characterization. The opening of the mouth and dental contact visible in the production of /ɨ/ are correlated with its acoustic measurements, albeit in a previously undocumented way. We conclude that Bora’s /ɨ/ should be reclassified as /i̪/ – front and dental.p { line-height: 115%; text-align: left; orphans: 2; widows: 2; margin-bottom: 0.1in; direction: ltr; background: transparent }

Parametrization of the vocal tract area function using a subset selection approach (L).

This letter introduces a parametrization of the vocal tract area function based on the position of a few points along the vocal tract. A QR decomposition algorithm is applied to area function data in various vowel configurations in order to identify those points with the most independent position patterns across vowels. Each point defines the shape of an associated kinematic region, and the overall area function is determined by the combination of the kinematic regions' shapes. The results show that only four data points, located at the tongue body, lips, and two at the tongue back, are enough to obtain accurate reconstructions of the vowels' area functions.

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.

Automatic detection of pharyngeal fricatives in cleft palate speech using acoustic features based on the vocal tract area spectrum

The pharyngeal fricative is a typical compensatory articulation disorder in cleft palate speech. It is produced by retracting the root of the tongue to the posterior pharyngeal wall to substitute for the fricatives and affricates produced in the oral cavity. People who use the pharyngeal fricative have difficulties in daily communication. Research on automatic pharyngeal fricative detection can provide aids in diagnosis for speech-language pathologists and clinical doctors. This work proposes a vocal tract area spectrum (VTAS) to represent a vocal tract model using time-varying cascaded pipes. Four acoustic features based on the VTAS (the centroid and spread (CS), peak linear deviation (PLD), relative-normal entropy (RNE), mean of the ratios’ statistics (MRS)) are proposed to evaluate the differences between pharyngeal fricatives and normal speech. The CS feature is proposed to evaluate the overall shape of the vocal tract to detect whether there are abnormal gestures or movements of the articulators in speech production. The PLD and RNE features focus on the variation and complexity of each vocal tube's area during the whole pronunciation process. The MRS feature is proposed to describe the continuity of the vocal tract. To evaluate the effectiveness of these four features, pharyngeal fricative detection experiments are conducted using a pharyngeal fricative dataset. This dataset contains 1246 speech samples spoken by 50 cleft palate patients and 50 normal speakers, covering all types of initial consonants in which the pharyngeal fricative usually occurs. The detection accuracy of the pharyngeal fricative using the CS, PLD, RNE and MRS feature ranges from 80.66% to 90.21%. When using the proposed CS+PLD+RNE+MRS feature, an accuracy of 95.18% can be achieved on the pharyngeal fricative dataset.

Electrical Modeling of Two Tube Vocal Tract for Voice Pathology Detection

Voice pathology is interrelated to a vocal folds complication, and the vocal tract area which is attached to vocal folds shows a variable shapes in case of voice pathology. This variable shapes are tested to make the division of healthy and pathological voices. So, Length and area of vocal tract are used for finding electrical parameters using transmission line theory by considering the vocal tract as a sequence of tubes. The proposed work is used voice database from the Saarbrucken Voice Database (SVD) and real-time data from the Maharishi Markandeshwar (M.M) hospital, Mullana. The proposed method achieves a higher accuracy of 91.7 percent with electrical parameters compared to 88.2 percent with the acoustic parameters and 85.3 percent with physical parameters in the detection of voice pathology.

Detection &amp; Classification of Voice Pathology using Electrical Circuit Parameters

The classification of pathological voice is a hot topic that has been expected significant consideration. Voice pathology is related with a vocal folds difficulty, and for this reason, the vocal tract area which is joined to vocal folds demonstrate random patterns in case of a pathological voice. This random pattern is considered to distinguish healthy and pathological voices. It is possible to utilize transmission line theory in discovering automatic voice pathology detection by taking into consideration the vocal tract as acoustic lines. The work concentrates on developing a feature extraction for detecting and classifying vocal fold polyp by investigating different vocal tract parameters. In this paper, the vocal tract length and area are utilized for computing electrical parameters of the vocal tract. Furthermore, these electrical parameters are used for the classification of pathological voice. Finally, using electrical parameters 97.3% accuracy is obtained with SVM classifier when compared with 88.2% with the acoustic parameters, 85.3% accuracy considering physical parameters and other methods used in the past. The outcomes demonstrate that electrical parameters of the vocal tract can be utilized all the more successfully with better precision in voice pathology identification.

Articulatory-acoustic relations in the production of alveolar and palatal lateral sounds in Brazilian Portuguese

Lateral approximant speech sounds are notoriously difficult to measure and describe due to their complex articulation and acoustics. This has prevented researchers from reaching a unifying description of the articulatory and acoustic characteristics of laterals. This paper examines articulatory and acoustic properties of Brazilian Portuguese alveolar and palatal lateral approximants (/l/ and /ʎ/) produced by six native speakers. The methodology for obtaining vocal tract area functions was based on three-dimensional/four-dimensional (3D/4D) ultrasound recordings and 3D digitized palatal impressions with simultaneously recorded audio signals. Area functions were used to calculate transfer function spectra, and predicted formant and anti-resonance frequencies were compared with the acoustic recordings. Mean absolute error in formant frequency prediction was 4% with a Pearson correlation of r = 0.987. Findings suggest anti-resonances from the interdental channels are less important than a prominent anti-resonance from the supralingual cavity but can become important in asymmetrical articulations. The use of 3D/4D ultrasound to study articulatory-acoustic relations is promising, but significant limitations remain and future work is needed to make better use of 3D/4D ultrasound data, e.g., by combining it with magnetic resonance imaging.

Optimal prosodic feature extraction and classification in parametric excitation source information for Indian language identification using neural network based Q-learning algorithm

Automatic language identification (LID) system has extensively recognized in a real world multilanguage speech specific applications. The formation speech is relying on the vocal tract area which explores the excitation source information for LID task. In this paper, LID system utilizes sub segmental, segmental and supra segmental features from Linear Prediction residual of speech signal, represents various native language speech excitation source information. The glottal flow derivative of speech signal is obtained through iterative adaptive inverse filtering method. Moreover, the prosodic features of speech signal are extracted using short time Fourier transform due to its capability to process non-stationary signals. Finally, the deep neural network based Q-learning (DNNQL) algorithm has been employed for identification of the class label for a specific language. Experimental validation of the proposed approach is carried out using Indian language recorded database. Finally, the proposed LID system approach is performing well with 97.3% accuracy compared to other machine learning based approaches.

Discrimination of “hot potato voice” caused by upper airway obstruction utilizing a support vector machine

"Hot potato voice" (HPV) is a thick, muffled voice caused by pharyngeal or laryngeal diseases characterized by severe upper airway obstruction, including acute epiglottitis and peritonsillitis. To develop a method for determining upper-airway emergency based on this important vocal feature, we investigated the acoustic characteristics of HPV using a physical, articulatory speech synthesis model. The results of the simulation were then applied to design a computerized recognition framework using a mel-frequency cepstral coefficient domain support vector machine (SVM). Quasi-experimental research design. Changes in the voice spectral envelope caused by upper airway obstructions were analyzed using a hybrid time-frequency model of articulatory speech synthesis. We evaluated variations in the formant structure and thresholds of critical vocal tract area functions that triggered HPV. The SVMs were trained using a dataset of 2,200 synthetic voice samples generated by an articulatory synthesizer. Voice classification experiments on test datasets of real patient voices were then performed. On phonation of the Japanese vowel /e/, the frequency of the second formant fell and coalesced with that of the first formant as the area function of the oropharynx decreased. Changes in higher-order formants varied according to constriction location. The highest accuracy afforded by the SVM classifier trained with synthetic data was 88.3%. HPV caused by upper airway obstruction has a highly characteristic spectral envelope. Based on this distinctive voice feature, our SVM classifier, who was trained using synthetic data, was able to diagnose upper-airway obstructions with a high degree of accuracy. 2c Laryngoscope, 129:1301-1307, 2019.

Automatic Detection Algorithm of Pharyngeal Fricatives in Cleft Palate Speech Based on LPIF and Feature Selection

Cleft palate is one of common congenital malformations that has huge impacts on the physical and psychological health of patients. Pharyngeal fricative in cleft palate speech is a kind of common compensatory articulations, which is produced by retracting tongue position to posterior pharyngeal wall and narrowing velopharyngeal opening. In this paper, based on voice mechanism of pharyngeal fricatives in cleft palate speech, linear prediction inverse filter was used to extract glottal waveform and estimate vocal tract model coefficients. Then four features were extracted, including pitch period, glottal flow derivative waveform, vocal tract area and vocal tract gain. The cross-correlation function was used to calculate the correlation between features. Glottal flow derivative waveform was removed since it had strong correlation with the others. KNN classifier was applied to realize the automatic pharyngeal fricatives detection in cleft palate speech, which reaches the detection accuracy of 98.40%.

3D dynamic MRI of the vocal tract during natural speech.

To develop and evaluate a technique for 3D dynamic MRI of the full vocal tract at high temporal resolution during natural speech. We demonstrate 2.4 × 2.4 × 5.8 mm3 spatial resolution, 61-ms temporal resolution, and a 200 × 200 × 70 mm3 FOV. The proposed method uses 3D gradient-echo imaging with a custom upper-airway coil, a minimum-phase slab excitation, stack-of-spirals readout, pseudo golden-angle view order in kx -ky , linear Cartesian order along kz , and spatiotemporal finite difference constrained reconstruction, with 13-fold acceleration. This technique is evaluated using in vivo vocal tract airway data from 2 healthy subjects acquired at 1.5T scanner, 1 with synchronized audio, with 2 tasks during production of natural speech, and via comparison with interleaved multislice 2D dynamic MRI. This technique captured known dynamics of vocal tract articulators during natural speech tasks including tongue gestures during the production of consonants "s" and "l" and of consonant-vowel syllables, and was additionally consistent with 2D dynamic MRI. Coordination of lingual (tongue) movements for consonants is demonstrated via volume-of-interest analysis. Vocal tract area function dynamics revealed critical lingual constriction events along the length of the vocal tract for consonants and vowels. We demonstrate feasibility of 3D dynamic MRI of the full vocal tract, with spatiotemporal resolution adequate to visualize lingual movements for consonants and vocal tact shaping during natural productions of consonant-vowel syllables, without requiring multiple repetitions.

Construction and Evaluation of a Parametric One-Dimensional Vocal Tract Model

Articulatory speech synthesis based on aero-acoustic simulations of the vocal tract is computationally expensive and, therefore, requires simple yet precise models. Modeling the one-dimensional vocal tract area function directly instead of a higher dimensional vocal tract model is an efficient way to minimize the computational overhead of the simulations. In this paper, we propose a new parametric vocal tract model that is controlled by six points and capable of modeling a large variety of vocal tract shapes. We geometrically and perceptually evaluated the model on a set of 22 reference area functions corresponding to German vowels and consonants. The model was able to geometrically approximate the reference area functions with a minimum root-mean-square error of 0.302 cm$^2$, a maximum error of 1.142 cm$^2$, and a median error of 0.891 cm$^2$. After optimizations, a perceptual evaluation of the synthesis using our model in combination with a state-of-the-art aero-acoustic simulation achieved a vowel recognition rate of 90.7% and a consonant recognition rate of 73.2%.

An age-dependent vocal tract model for males and females based on anatomic measurements.

The purpose of this study was to take a first step toward constructing a developmental and sex-specific version of a parametric vocal tract area function model representative of male and female vocal tracts ranging in age from infancy to 12 yrs, as well as adults. Anatomic measurements collected from a large imaging database of male and female children and adults provided the dataset from which length warping and cross-dimension scaling functions were derived, and applied to the adult-based vocal tract model to project it backward along an age continuum. The resulting model was assessed qualitatively by projecting hypothetical vocal tract shapes onto midsagittal images from the cohort of children, and quantitatively by comparison of formant frequencies produced by the model to those reported in the literature. An additional validation of modeled vocal tract shapes was made possible by comparison to cross-sectional area measurements obtained for children and adults using acoustic pharyngometry. This initial attempt to generate a sex-specific developmental vocal tract model paves a path to study the relation of vocal tract dimensions to documented prepubertal acoustic differences.

Determining the shape of a human vocal tract from pressure measurements at the lips

The inverse problem of determining the cross-sectional area of a human vocal tract during the utterance of a vowel is considered in terms of the data consisting of the absolute value of sound pressure at the lips. If the upper lip is curved downward during the utterance, it is shown that there may be up to an M-fold nonuniqueness in the determination, where M is the maximal number of eligible resonances associated with a related Schrödinger operator. Each of the M such distinct candidates for the vocal-tract area corresponding to the same absolute pressure is uniquely determined. The mathematical theory is presented for the recovery of each candidate for the vocal-tract area, and the admissibility criterion for each of the M candidates to be a vocal-tract radius is specified. On the other hand, if the upper lip is horizontal or curved upward during the utterance, then the inverse problem has a unique solution. The theory developed is illustrated with some examples.

Articulatory correlates of phonemic and coarticulatory nasalization

Phonological theory distinguishes nasal and oral vowel counterparts by velopharyngeal port opening, neglecting other phonetic differences between phonemic and coarticulatory nasalization. Recent articulatory work provides evidence of oropharyngeal distinctions, in addition to velic lowering. This study (12 Brazilian Portuguese speakers) uses real-time MRI to investigate oropharyngeal differences between oral, phonemically nasal, and phonetically nasalized vowels /a, i, u/. Tissue boundaries in midsagittal vocal tract images were automatically detected to reveal each vowel repetition’s aperture function. Principal Components Analysis determined vocal tract regions responsible for the greatest variance in the data. Time-dynamic analyses of vocal tract area in these regions used smoothing spline ANOVA. Results show the tongue body and/or hyperpharynx as the most important articulators. For /a/, nasal vowels demonstrate wider hyperpharyngeal and narrower tongue body regions compared to oral vowels. For /u/, o...

Development of Visual Support Way for Pronunciation Practice Using Reflection Coefficients of Burg's Method

In pronunciation training system, it is effective to instruct the manner of articulation to the trainee. We have developed a pronunciation training system, where the lip posture and tongue shape that are estimated by the formants extracted by AR model using Burg's method are shown to the trainee. However, formants estimation requires long computation time. In this report, we attempt to reconstruct these visual instructions from the reflection coefficients of Burg's method. As a results of principal component analysis for the reflection coefficients, it was shown that the discrimination of the vowels only by the reflection coefficients is difficult. On the other hand, the tongue shape, the front and back position of the tongue, was effectively visualized using vocal tract area function estimated by the reflection coefficients.

An acoustically-driven vocal tract model for stop consonant production

The purpose of this study was to further develop a multi-tier model of the vocal tract area function in which the modulations of shape to produce speech are generated by the product of a vowel substrate and a consonant superposition function. The new approach consists of specifying input parameters for a target consonant as a set of directional changes in the resonance frequencies of the vowel substrate. Using calculations of acoustic sensitivity functions, these “resonance deflection patterns” are transformed into time-varying deformations of the vocal tract shape without any direct specification of location or extent of the consonant constriction along the vocal tract. The configuration of the constrictions and expansions that are generated by this process were shown to be physiologically-realistic and produce speech sounds that are easily identifiable as the target consonants. This model is a useful enhancement for area function-based synthesis and can serve as a tool for understanding how the vocal tract is shaped by a talker during speech production.

Robust phoneme classification for automatic speech recognition using hybrid features and an amalgamated learning model

Phoneme recognition is an important aspect of speech processing and recognition. Research on phoneme recognition is several years old and numerous algorithms have been developed over the years to improve its accuracy. In this paper, a quantitative analysis of phoneme recognition using supervised learning is investigated. Most approaches to phoneme recognition rely on using mel frequency cepstrum based features for identification of the phoneme class. In our approach, we take into consideration the vocal tract area function along with mel frequency cepstrum coefficients and analyze the change in accuracy obtained by its introduction in the feature set. Support Vector Machines have been an attractive approach to pattern recognition and its usage as a supervised learning model has been popular in the speech processing community. We compare Support Vector Machines to other supervised learning models like the Naive Bayes, the k-Nearest Neighbors and the linear discriminant analysis classifiers, for our feature set. We impose a soft voting rule between the three best classifiers to produce our variation of a voting classifier. We enhance the accuracy of our classifier by using a priority based approach to estimate the three most likely phonemes, after the predicted phoneme. Through a figurative and quantitative approach, we show that our modified algorithm outperforms other traditional methods. Experiments were conducted on the WSJCAM0 corpus, a British English corpus.