Voice Source Research Articles

Modeling the voice source in terms of spectral slopes

A psychoacoustic model of the voice source spectrum is proposed. The model is characterized by four spectral slope parameters: the difference in amplitude between the first two harmonics (H1-H2), the second and fourth harmonics (H2-H4), the fourth harmonic and the harmonic nearest 2 kHz in frequency (H4-2 kHz), and the harmonic nearest 2 kHz and that nearest 5 kHz (2 kHz-5 kHz). As a step toward model validation, experiments were conducted to establish the acoustic and perceptual independence of these parameters. In experiment 1, the model was fit to a large number of voice sources. Results showed that parameters are predictable from one another, but that these relationships are due to overall spectral roll-off. Two additional experiments addressed the perceptual independence of the source parameters. Listener sensitivity to H1-H2, H2-H4, and H4-2 kHz did not change as a function of the slope of an adjacent component, suggesting that sensitivity to these components is robust. Listener sensitivity to changes in spectral slope from 2 kHz to 5 kHz depended on complex interactions between spectral slope, spectral noise levels, and H4-2 kHz. It is concluded that the four parameters represent non-redundant acoustic and perceptual aspects of voice quality.

Segmentation of the period of the fundamental tone of a voice source

The extrema of the logarithmic derivative of the mean energy of a voice signal in the frequency range of 1000–3000 Hz are used to determine the instants of opening and closure of the glottis. The inaccuracy of analysis is estimated with the Arctic CMU database, which contains synchronous recordings of speech signals and electro-glottograms. The estimates of the instants of opening and closure of the glottis, found by the developed algorithm, are compared with the instants of the maximum and minimum of the derivative from electro-glottogram signals, which are taken as the “true” instants. The mean square deviation of the glottal opening instant from the extrema of the derivative from the electro-glottogram signals for different speakers is in the range of 1.03–1.64 ms. The error rate of a false estimate of the glottal opening instant is from 0.01 to 0.14%, and the error rate of omission is from 0.42 to 2.38%. An error-detection algorithm is developed. The mean square deviation with an relative—to the period of the fundamental tone—error in detecting the glottal opening instant is in the range of 13–18% for the most probable error from 0 to +5%.

Acoustic Properties of the Voice Source and the Vocal Tract: Are They Perceptually Independent?

This study sought to determine whether the properties of the voice source and vocal tract are perceptually independent. Within-subjects design. This study employed a paired-comparison paradigm where listeners heard synthetic voices and rated them as same or different using a visual analog scale. Stimuli were synthesized using three different source slopes and two different formant patterns (mezzo-soprano and soprano) on the vowel /a/ at four pitches: A3, C4, B4, and F5. Whereas formant pattern was the strongest effect, difference in source slope also affected perceived quality difference. Source slope and formant pattern were not independently perceived. These results suggest that when judging laryngeal adduction using perceptual information, judgments may not be accurate when the stimuli are of differing formant patterns.

Detection of Voice Pathology using Fractal Dimension in a Multiresolution Analysis of Normal and Disordered Speech Signals.

Voice disorders are associated with irregular vibrations of vocal folds. Based on the source filter theory of speech production, these irregular vibrations can be detected in a non-invasive way by analyzing the speech signal. In this paper we present a multiband approach for the detection of voice disorders given that the voice source generally interacts with the vocal tract in a non-linear way. In normal phonation, and assuming sustained phonation of a vowel, the lower frequencies of speech are heavily source dependent due to the low frequency glottal formant, while the higher frequencies are less dependent on the source signal. During abnormal phonation, this is still a valid, but turbulent noise of source, because of the irregular vibration, affects also higher frequencies. Motivated by such a model, we suggest a multiband approach based on a three-level discrete wavelet transformation (DWT) and in each band the fractal dimension (FD) of the estimated power spectrum is estimated. The experiments suggest that frequency band 1-1562 Hz, lower frequencies after level 3, exhibits a significant difference in the spectrum of a normal and pathological subject. With this band, a detection rate of 91.28 % is obtained with one feature, and the obtained result is higher than all other frequency bands. Moreover, an accuracy of 92.45 % and an area under receiver operating characteristic curve (AUC) of 95.06 % is acquired when the FD of all levels is fused. Likewise, when the FD of all levels is combined with 22 Multi-Dimensional Voice Program (MDVP) parameters, an improvement of 2.26 % in accuracy and 1.45 % in AUC is observed.

Contrôle gestuel de la synthèse vocale. Les instruments Cantor Digitalis et Digitartic

EXTENDED ABSTRACT: Gestural control of speech or singing synthesis is difficult, because of the very fast articulators motions encountered in speech and singing. For singing, another difficult question is accurate rhythmic coordination and precision, because syllables must coincide with musical beats, at a given tempo. Then, the precise location of beats for different syllables must be controlled. Two singing synthesis instruments are presented: Digitalis Cantor and Digitartic. Both instruments use bimanual writing or drawing gestures on graphic tablets. The voice signal is computed with the help of a parametric synthesizer, including a voice source model, consonantal noise models and series/parallel formant filters. Cantor Digitalis is a vowel and semi-vowel singing instrument. Digitartic is an extension of Cantor Digitalis and allows for singing syllables, including plosives, fricative, liquid and nasal consonants. Any in-between canonical place of articulation is possible by linear interpolation of the consonant parameters, for each mode of articulation. In this paper, the focus is given on Digitartic through the issue of consonant gestures and musical beat synchronization. Three modes of Vowel-Consonant-Vowel (VCV) articulation are discussed according to three levels of rhythmic precision and musical context. A VCV articulation is composed of the onset phase (articulators approaching the position of maximum constriction), the medial phase (maximum of constriction) and the offset phase (constriction release). Offset phase of plosives is very short compared to other consonants. The first control mode consists of triggering the syllable at the beginning of the onset phase. However, when the syllable starts on the musical beat, it is perceived with a delay depending on the duration of articulation phases. Anticipating precisely this delay is very difficult. The second control mode of control allows for triggering the VCV dissyllable in two steps. In the first step, the onset phase is triggered, and in the second step, the offset phase is triggered. In this way, plosives can be synchronized with musical beats without any delay. The third control mode is a continuous control of the phases of articulation, without any triggering. This requires a fast synthesis engine, a high interface sampling rate, as well as an expert control gesture, fast and precise enough to reproduce speech articulation phases. The continuous control mode of articulation is performed by a back-and-forth gesture with the pen of the non-preferred hand, along the vertical dimension of the graphic tablet. Place of articulation is continuously controlled along the horizontal dimension, and the mode of articulation is assigned to different areas on the tablet. This back-and-forth gesture is analog to the somewhat symmetric articulation of the VCV dissyllable. The gesture amplitude allows for different degrees of articulation (hypoarticulation to hyperarticulation). Controlling durations of each phase of articulation is another mean to increase expressiveness. The preferred hand is controlling pitch, vocal effort and vowel quality on another graphic tablet. Then it is possible to modify pitch and vocal effort during each phase of articulation. Cantor Digitalis and Digitartic allow for expressive musical performances. They are regularly used for concerts.

Analysis of Measured and Simulated Supraglottal Acoustic Waves

To date, although much attention has been paid to the estimation and modeling of the voice source (ie, the glottal airflow volume velocity), the measurement and characterization of the supraglottal pressure wave have been much less studied. Some previous results have unveiled that the supraglottal pressure wave has some spectral resonances similar to those of the voice pressure wave. This makes the supraglottal wave partially intelligible. Although the explanation for such effect seems to be clearly related to the reflected pressure wave traveling upstream along the vocal tract, the influence that nonlinear source-filter interaction has on it is not as clear. This article provides an insight into this issue by comparing the acoustic analyses of measured and simulated supraglottal and voice waves. Simulations have been performed using a high-dimensional discrete vocal fold model. Results of such comparative analysis indicate that spectral resonances in the supraglottal wave are mainly caused by the regressive pressure wave that travels upstream along the vocal tract and not by source-tract interaction. On the contrary and according to simulation results, source-tract interaction has a role in the loss of intelligibility that happens in the supraglottal wave with respect to the voice wave. This loss of intelligibility mainly corresponds to spectral differences for frequencies above 1500Hz.

Voice Source Variation Between Vowels in Male Opera Singers

The theory of nonlinear source-filter interaction predicts that the glottal voice source should be affected by the frequency relationship between formants and partials. An attempt to experimentally verify this theory is presented. Glottal voice source and electrolaryngograph (ELG) signal differences between vowels were analyzed in vowel sequences, sung at four pitches with the same degree of vocal loudness by professional opera singers. In addition, the relationships between such differences and the frequency distance between the first formant (F1) and its closest partial were examined. A digital laryngograph microprocessor was used to simultaneously record audio and ELG signals. The former was inverse filtered, and voice source parameters and formant frequencies were extracted. The amplitude quotient of the derivative of the ELG signal (AQdELG) and the contact quotient were also compared. A one-way repeated-measures ANOVA revealed significant differences between vowels, for contact quotient at four pitches and for maximum flow declination rate (MFDR) at three pitches. For other voice source parameters, differences were found at one or two pitches only. No consistent correlation was found between MFDR and the distance between F1 and its closest partial. The glottal voice source tends to vary between vowels, presumably because of source-filter interaction, but the variation does not seem to be dependent on the frequency distance between F1 and its closest partial.

Perceptual evaluation of voice source models.

Models of the voice source differ in their fits to natural voices, but it is unclear which differences in fit are perceptually salient. This study examined the relationship between the fit of five voice source models to 40 natural voices, and the degree of perceptual match among stimuli synthesized with each of the modeled sources. Listeners completed a visual sort-and-rate task to compare versions of each voice created with the different source models, and the results were analyzed using multidimensional scaling. Neither fits to pulse shapes nor fits to landmark points on the pulses predicted observed differences in quality. Further, the source models fit the opening phase of the glottal pulses better than they fit the closing phase, but at the same time similarity in quality was better predicted by the timing and amplitude of the negative peak of the flow derivative (part of the closing phase) than by the timing and/or amplitude of peak glottal opening. Results indicate that simply knowing how (or how well) a particular source model fits or does not fit a target source pulse in the time domain provides little insight into what aspects of the voice source are important to listeners.

Voice source characterization using pitch synchronous discrete cosine transform for speaker identification.

A characterization of the voice source (VS) signal by the pitch synchronous (PS) discrete cosine transform (DCT) is proposed. With the integrated linear prediction residual (ILPR) as the VS estimate, the PS DCT of the ILPR is evaluated as a feature vector for speaker identification (SID). On TIMIT and YOHO databases, using a Gaussian mixture model (GMM)-based classifier, it performs on par with existing VS-based features. On the NIST 2003 database, fusion with a GMM-based classifier using MFCC features improves the identification accuracy by 12% in absolute terms, proving that the proposed characterization has good promise as a feature for SID studies.

Voice source under cognitive load: Effects and classification

Cognitive load refers to the mental demand experienced while performing a cognitive task. A cognitive load measurement system can potentially be a useful tool for monitoring and enhancing human task performance. In the area of speech-based cognitive load classification, while there are various spectral and vocal tract-based features proposed for classification purposes, there is still a lack of studies that investigate how cognitive load affects the voice source, and whether glottal features are effective in cognitive load classification systems. This work introduces a set of databases that contains both speech and electroglottograph (EGG) data. Using these databases, we present results that provide arguably the first direct insight into how cognitive load affects the voice source. Additionally, we show that glottal-based features carry complementary information with respect to formant-based features, and that fusion between glottal and formant-based systems produces classification results that are comparable with (if not better than) existing baseline systems across three out of five evaluation databases.

Flow Glottogram Characteristics and Perceived Degree of Phonatory Pressedness

Phonatory pressedness is a clinically relevant aspect of voice, which generally is analyzed by auditory perception. The present investigation aimed at identifying voice source and formant characteristics related to experts' ratings of phonatory pressedness. Experimental study of the relations between visual analog scale ratings of phonatory pressedness and voice source parameters in healthy voices. Audio, electroglottogram, and subglottal pressure, estimated from oral pressure during /p/ occlusion, were recorded from five female and six male subjects, each of whom deliberately varied phonation type between neutral, flow, and pressed in the syllable /pae/, produced at three loudness levels and three pitches. Speech-language pathologists rated, along a visual analog scale, the degree of perceived phonatory pressedness in these samples. The samples were analyzed by means of inverse filtering with regard to closed quotient, dominance of the voice source fundamental, normalized amplitude quotient, peak-to-peak flow amplitude, as well as formant frequencies and the alpha ratio of spectrum energy above and below 1000Hz. The results were compared with the rating data, which showed that the ratings were closely related to voice source parameters. Approximately, 70% of the variance of the ratings could be explained by the voice source parameters. A multiple linear regression analysis suggested that perceived phonatory pressedness is related most closely to subglottal pressure, closed quotient, and the two lowest formants.

Glottal Adduction and Subglottal Pressure in Singing

Previous research suggests that independent variation of vocal loudness and glottal configuration (type and degree of vocal fold adduction) does not occur in untrained speech production. This study investigated whether these factors can be varied independently in trained singing and how subglottal pressure is related to average glottal airflow, voice source properties, and sound level under these conditions. A classically trained baritone produced sustained phonations on the endoscopic vowel [i:] at pitch D4 (approximately 294 Hz), exclusively varying either (a) vocal register; (b) phonation type (from "breathy" to "pressed" via cartilaginous adduction); or (c) vocal loudness, while keeping the others constant. Phonation was documented by simultaneous recording of videokymographic, electroglottographic, airflow and voice source data, and by percutaneous measurement of relative subglottal pressure. Register shifts were clearly marked in the electroglottographic wavegram display. Compared with chest register, falsetto was produced with greater pulse amplitude of the glottal flow, H1-H2, mean airflow, and with lower maximum flow declination rate (MFDR), subglottal pressure, and sound pressure. Shifts of phonation type (breathy/flow/neutral/pressed) induced comparable systematic changes. Increase of vocal loudness resulted in increased subglottal pressure, average flow, sound pressure, MFDR, glottal flow pulse amplitude, and H1-H2. When changing either vocal register or phonation type, subglottal pressure and mean airflow showed an inverse relationship, that is, variation of glottal flow resistance. The direct relation between subglottal pressure and airflow when varying only vocal loudness demonstrated independent control of vocal loudness and glottal configuration. Achieving such independent control of phonatory control parameters would be an important target in vocal pedagogy and in voice therapy.

Articulation and vocal tract acoustics at soprano subject's high fundamental frequencies.

The role of the vocal tract for phonation at very high soprano fundamental frequencies (F0s) is not yet understood in detail. In this investigation, two experiments were carried out with a single professional high soprano subject. First, using two dimensional (2D) dynamic real-time magnetic resonance imaging (MRI) (24 fps) midsagittal and coronal vocal tract shapes were analyzed while the subject sang a scale from Bb5 (932 Hz) to G6 (1568 Hz). In a second experiment, volumetric vocal tract MRI data were recorded from sustained phonations (13 s) for the pitches C6 (1047 Hz) and G6 (1568 Hz). Formant frequencies were measured in physical models created by 3D printing, and calculated from area functions obtained from the 3D vocal tract shapes. The data showed that there were only minor modifications of the vocal tract shape. These changes involved a decrease of the piriform sinus as well as small changes of tongue position. Formant frequencies did not exhibit major differences between C6 and G6 for F1 and F3, respectively. Only F2 was slightly raised for G6. For G6, however, F2 is not excited by any voice source partial. Therefore, this investigation was not able to confirm that the analyzed professional soprano subject adjusted formants to voice source partials for the analyzed F0s.

Sufficiency of a four-parameter spectral model of the voice source

Our proposed perceptually motivated spectral-domain model of the voice source comprises spectral slopes in four ranges (H1-H2, H2-H4, H4-the harmonic nearest 2 kHz, and that harmonic to the harmonic nearest 5 kHz). Previous studies established the necessity of these parameters by demonstrating that listeners are sensitive to all of these parameters, and that all are needed to adequately model voice quality. The present study examines the sufficiency of the model to quantify voice quality by applying it to a very large set of voices. To determine the range of phenomena for which the source model is valid, 200 natural voices (half male; 50 normal/mildly pathologic, 30 moderately pathologic, 20 severely pathologic) were copy-synthesized using the model, and adequacy of synthesis was assessed in a listening test by comparing the synthetic tokens to the natural targets. Validity of the model as a function of severity of pathology and the particular manner in which the stimuli deviate from normal will be discus...

Ken Stevens’ influence on the development of paradigms for speech synthesis

Synthetic speech has long been used as a means of understanding both speech production and speech perception, as well as for technological applications such as text-to-speech devices. Paradigms for developing speech synthesis systems have included electrical circuits, digital filters, and computational models that replicate either the structure or acoustic characteristics of the voice source and vocal tract. This presentation will focus on how Ken Stevens’ investigations of speech, spanning more than five decades, have directly influenced essentially every paradigm of speech synthesis, including formant synthesis, articulatory synthesis, and speech production modeling. [Work supported by NIH R01-DC011275 and NSF BCS-1145011.]

Respiratory and Acoustical Differences Between Belt and Neutral Style of Singing

Belt is a style of singing commonly used in nonclassical genres. Its respiratory, phonatory, and resonatory characteristics are unclear. Basic research. Six female singers, professionally performing in the belt styles since many years, sang an excerpt of a song in belt and nonbelt/neutral style, two times with the lyrics and two times replacing the lyrics with /pae/ syllables. On separate channels, recordings were made of audio, oral pressure, and rib cage and abdominal wall movements, as picked up by respiratory inductive plethysmography. Lung volume and breathing patterns during inhalation and phonation were normalized with respect to duration and averaged. Voice source was analyzed in terms of flow glottograms derived from the audio signal by inverse filtering. Belt was produced with higher pressures and yielded higher sound levels, but no consistent breathing pattern was observed, neither for the belt, nor for the neutral style. Voice source differences suggested that belt was produced with firmer glottal adduction than neutral. Also, in four of the singers, the first formant was closer to a spectrum harmonic in belt than in neutral. Belt style of singing is not associated with a characteristic breathing behavior but is produced with higher subglottal pressures, higher sound levels, and firmer glottal adduction than a neutral style of singing.

Perception of glottalization and phrase-final creak.

American English has several linguistic sources of creaky voice. Two common sources are /t/-glottalization (where /t/ is produced as a glottal stop and/or with creaky voice, as in "button") and phrase-final creak. Both /t/-glottalization and phrase-final creak have similar acoustic properties, but they can co-occur in English. The goal of this study is to determine whether /t/-glottalization and phrase-final creak are perceived distinctly. Sixteen English listeners were asked to identify words in a two-alternative forced choice task. The auditory targets were (near-) minimal pairs, in which one word could have /t/-glottalization (e.g., "button") but the other could not (e.g., "bun"). Stimuli were presented with and without phrase-final creak. Listeners made few identification errors overall, even when /t/-glottalization co-occurred with phrase-final creak, suggesting that /t/-glottalization and phrase-final creak remain perceptually distinct to English listeners. This supports the view that creaky voice is not a single category, but one comprised of distinct voice qualities.

Computerized analysis of acoustic characteristics of patients with internal nasal valve collapse before and after functional rhinoplasty

Acoustic analysis of sounds produced during speech provides significant information about the physiology of larynx and vocal tract. The analysis of voice power spectrum is a fundamental sensitive method of acoustic assessment that provides valuable information about the voice source and characteristics of vocal tract resonance cavities. The changes in long-term average spectrum (LTAS) spectral tilt and harmony to noise ratio (HNR) were analyzed to assess the voice quality before and after functional rhinoplasty in patients with internal nasal valve collapse. Before and 3 months after functional rhinoplasty, 12 participants were evaluated and HNR and LTAS spectral tilt in /a/ and /i/ vowels were estimated. It was seen that an increase in HNR and a decrease in LTAS spectral tilt existed after surgery. Mean LTAS spectral tilt in vowel /a/ decreased from 2.37 ± 1.04 to 2.28 ± 1.17 (P = 0.388), and it was decreased from 4.16 ± 1.65 to 2.73 ± 0.69 in vowel /i/ (P = 0.008). Mean HNR in the vowel /a/ increased from 20.71 ± 3.93 to 25.06 ± 2.67 (P = 0.002), and it was increased from 21.28 ± 4.11 to 25.26 ± 3.94 in vowel /i/ (P = 0.002). Modification of the vocal tract caused the vocal cords to close sufficiently, and this showed that although rhinoplasty did not affect the larynx directly, it changes the structure of the vocal tract and consequently the resonance of voice production. The aim of this study was to investigate the changes in voice parameters after functional rhinoplasty in patients with internal nasal valve collapse by computerized analysis of acoustic characteristics.

Contact Quotient Versus Closed Quotient: A Comparative Study on Professional Male Singers

The term "closed quotient" is frequently used for data derived both from inverse filtering and from electroglottography. In the former case, it is defined as the ratio between the closed phase and the period, as measured in flow glottograms (FLOGG), whereas in the latter case, it is defined as the time interval between the falling and rising parts of the electroglottogram (EGG), measured at some percentage of the peak-to-peak amplitude. The study aims at analyzing differences between EGG- and FLOGG-based closed quotients and their relationships with voice source parameters. Comparative study. FLOGG- and EGG-based measures collected from five professional male singers were compared, under different pitch and loudness conditions. Compared with the FLOGG-based quotient, the EGG-based quotient (i) varied more between subjects, (ii) presented greater values, (iii) varied less with subglottal pressure, (iv) varied less with the normalized amplitude quotient (ie, the ratio between the flow pulse amplitude and the product of period and maximum flow declination rate), and (v) varied less with the relative amplitude of the voice source fundamental. Although positively related, FLOGG- and EGG-based closed quotients differ and must not be confused.

Experimental demonstration of differential OFDM underwater acoustic communication with acoustic vector sensor

A new Orthogonal Frequency Division Multiplexing (OFDM) based underwater acoustic communication scheme utilizing acoustic vector sensor is presented. The acoustic channel assumed has the slow time-varying characteristic and the neighboring signals can be considered as transmitted through the same channel. The differential modulation encodes information through the relative phase changes between adjacent subcarriers. Therefore, instead of adding pilots to monitor the channel, the differential demodulation is used to reconstructs transmitted information without necessity of channel estimation. This increases the transmission rate and reduce the complexity of processor at the receiver. An acoustic vector sensor is used in the differential OFDM system to further improve the performance of the system with 4.8dB gains that can be obtained by combined processing pressure and velocity signals from the acoustic vector sensor. To validate the capability of the proposed scheme, an experiment with picture and voice source information was conducted in a lake. The results obtained at different distances demonstrate the proposed system offers a significant advantage.