Non-modal Voice Quality Research Articles

Non-modal voice quality in Chilean Spanish

Abstract Since the turn of the 21st century, research on phonation (the production of sound via the vibration of the vocal folds, also known as voice quality) has increased significantly. However, non-modal phonation is highly understudied in Spanish. Therefore, this study comprises a sociophonetic analysis of voice quality in Spanish, examining voice quality categorically from sociolinguistic interview speech conducted with 41 native monolingual Chilean Spanish speakers. The aim of the present paper is to interrogate how phonation varies among residents of Santiago, Chile. Specifically, it asks: what is the sociolinguistic distribution of non-modal voice quality in Chilean Spanish? How does this distribution align with work on voice quality in Spanish and in other languages? Results indicate that non-modal voice quality is used by speakers to mark prosodic boundaries such as phrase finality, aligning with robust cross- linguistic findings. Non-modal phonation is also incorporated when speakers are reporting speech, as well as when they are speaking about emotionally-charged subjects. I draw connections between these findings and those of other dialects of Spanish, and show that utterance-final devoicing, while often associated with “tierras altas” dialects of Spanish, may be more frequent in other dialects than previously indicated. I argue that further examinations of non-modal phonation should also examine third-wave approaches of attitude, emotion, and stance.

The prosodic marking of rhetorical questions in Standard Chinese

The present study investigates the prosody of information-seeking (ISQs) and rhetorical questions (RQs) in Standard Chinese, in polar and wh-questions. Like in other languages, ISQs and RQs in Standard Chinese can have the same surface structure, allowing for a direct prosodic comparison between illocution types (ISQ vs RQ). Since Standard Chinese has lexical tone, the use of f0 as a cue to illocution type may be restricted. We investigate the prosodic differences between ISQs and RQs as well as the interplay of prosodic cues to RQs. In terms of f0, results showed that RQs were lower in f0, with the f0 range on the first word being expanded followed by f0 compression. RQs were further longer in duration and more often realized with non-modal voice quality (glottalized voice) as compared to ISQs. These prosodic cues were largely manipulated in tandem (illocutionary pairs with larger durational differences also showed larger differences in mean f0; voice quality, in turn, seemed to be an additional cue). We suggest three possible explanations (assertive force, focus, speaker attitude) that unite the present findings on RQs in Standard Chinese with the findings on RQs in other, non-tonal languages.

Evaluating automatic creaky voice detection methods.

There is growing interest in research on the non-modal voice quality, creaky voice; however, its identification often relies on time-consuming manual annotation, leading to a recent focus on automatic creak detection methods. Various automatic methods have been proposed, which rely on varying types and combinations of acoustic cues for creak detection. In this paper, we compare the performance of three automatic tools, the AntiMode method, the Creak Detector algorithm, and the Roughness algorithm, against manual annotation of creak using data from 80 Australian English speakers. We explore the possibility that tools used in combination may yield more accurate creak detection than individual tools used alone. Based on method comparisons, we present options for researchers, including an "out-of-the-box" approach, which supports combining automatic tools, and propose additional steps to further improve creak detection. We found restricting analysis to sonorant segments significantly improves automatic creak detection, and tools performed consistently better on female speech than male speech. Findings support previous work showing detection may be optimised by performing a creak probability threshold sweep on a subset of data prior to applying the Creak Detector algorithm on new datasets. Results provide promising solutions for advancing efficient large-scale research on creaky voice.

Beyond binary gender: creaky voice, gender, and the variationist enterprise

AbstractThis paper promotes a sophisticated treatment of gender in variationism through a large-scale quantitative analysis of creak, a nonmodal voice quality stereotypically associated with women in US English. An analysis of our gender-diverse corpus, including cisgender, transgender, and nonbinary individuals, finds that gender does not predict variation; all gender groups produce high rates of creak. However, gender does interact with style: all speakers use more creak in interview speech compared with read speech, but some groups style-shift more than others, suggesting that gender remains a relevant factor in capturing how creak is deployed as a resource in social practice. We use this analysis to advocate for a move beyond the gender binary in quantitative descriptions of sociolinguistic variables and call for the greater inclusion of trans+ individuals in sociolinguistics.

Voice quality and speaking rate in Icelandic rhetorical questions

AbstractIn this paper, we show that Icelandic uses the phonetic parameters of speaking rate, duration and voice quality (VQ) to distinguish between information-seeking questions (ISQs) and rhetorical questions (RQs). Specifically, durations are longer (speaking rate is slower) and nonmodal VQs are used more in RQs than in ISQs. Our findings for temporal parameters fit in with previous studies on the prosody of RQs in various languages. With respect to VQ, Icelandic differs, for example, from German and English in the location of breathy voice in the utterance (utterance-initial in German and English, utterance-final in Icelandic). We interpret the utterance-final position of breathiness in Icelandic RQs as a potential compensating strategy for the lack of phonological cues, i.e. boundary tones.

The prosody of rhetorical questions: a cross-linguistic view

This paper provides a survey of our knowledge of the prosody of rhetorical questions, i.e. questions that do not require an answer and try to commit the listener to the presupposed answer, as compared to the prosody of string-identical genuine, information seeking-questions. The survey includes semantic literature on questions, corpus data, and experimental evidence from production and perception experiments. It covers a range of typologically different languages (German, English, Icelandic, Italian, Standard Chinese, Cantonese, Japanese, French) that have different word-level and phrase-level characteristics. The main finding is that rhetorical and information seeking questions differ reliably in terms of the following prosodic characteristics: (i) f0-features (e.g., position and type of pitch accent, type of boundary tone, as well as more global f0-parameters, depending on language type); (ii) duration / speaking rate (rhetorical questions are typically longer / produced with slower speaking rate than information-seeking questions). Often, but not always, rhetorical questions are produced with non-modal voice quality.

Classifying females\u2019 stressed and neutral voices using acoustic\u2013phonetic analysis of vowels: an exploratory investigation with emergency calls

In the present exploratory study, we investigated acoustic–phonetic measures of spoken vowels for detection of female speech under conditions of stress. Eight authentic recorded calls to emergency services received from eight Finnish adult female speakers were chosen for the analysis. Based on the purpose of the call, the recordings were divided into two groups: the stressed group and the neutral group. We chose f0, H1–H2 and centre of gravity as acoustic–phonetic predictors for our final classification models; In previous studies, high f0 has been associated with a stressed voice, but H1–H2 and centre of gravity have not previously been related to speech under stress. On the other hand, H1–H2 has been used to detect non-modal voice qualities, such as a creaky or breathy voice, and similar voice qualities have been observed in stressed speech. Furthermore, indications exist that in speech under stress, acoustic energy is concentrated in higher frequencies, which consequently increases the centre of gravity. We tested stress detection accuracy with three statistical classifiers: LDA, logistic regression and decision tree. Our results indicated that all the models performed better when they were trained using only the vowel /i/ rather than training them with all Finnish vowels. The use of our best performing model, a logistic regression model based on /i/, yielded 88% correct classification, whereas a logistic regression model trained with all vowels achieved an accuracy of only 81%. We conclude that the results indicate a good stress classification accuracy, although further research with more extensive data is required.

Cross-language differences in how voice quality and f 0 contours map to affect.

The relationship between prosody and perceived affect involves multiple variables. This paper explores the interplay of three: voice quality, f 0 contour, and the hearer's language background. Perception tests were conducted with speakers of Irish English, Russian, Spanish, and Japanese using three types of synthetic stimuli: (1) stimuli varied in voice quality, (2) stimuli of uniform (modal) voice quality incorporating affect-related f 0 contours, and (3) stimuli combining specific non-modal voice qualities with the affect-related f 0 contours of (2). The participants rated the stimuli for the presence/strength of affective colouring on six bipolar scales, e.g., happy-sad. The results suggest that stimuli incorporating non-modal voice qualities, with or without f 0 variation, are generally more effective in affect cueing than stimuli varying only in f 0. Along with similarities in the affective responses across these languages, many points of divergence were found, both in terms of the range and strength of affective responses overall and in terms of specific stimulus-to-affect associations. The f 0 contour may play a more important role, and tense voice a lesser role in affect signalling in Japanese and Spanish than in Irish English and Russian. The greatest cross-language differences emerged for the affects intimate, formal, stressed, and relaxed.

Duration and Voice Quality of Early Infant Vocalizations.

Infant vocal durations have been studied from a variety of perspectives, including medical, social, and linguistic. The resultant developmental profile across the first 6 months of life, however, is still far from clear. The purpose of this study was to evaluate the durational properties of infant vocalizations from the unique perspective of voice quality. By considering an infant's modal and nonmodal voice qualities, the developmental range of vocalizations produced by infants during the early months of life was captured. Four Australian English-speaking infants were recorded for approximately 1 hr per week during the first 6 months of life. A total of 6,309 vocalizations were perceptually identified and labeled according to voice quality. The duration of each vocalization was subsequently measured. A nonlinear curve was evident for the duration of all vocalizations combined. Duration increased significantly between Months 3 and 5. Modal voice was the only voice quality that displayed a linear increase in duration across the study. All other voice qualities displayed polynomial trends. Based on the current results, the inconsistent pattern of vocal duration development found previously can be reconciled when voice quality properties of vocalizations are taken into account. A nonlinear curve is evident when a broad corpus of infant vocalizations is used, whereas a narrow corpus containing predominantly modal vocalizations displays a linear trend. The results demonstrate the necessity of including nonmodal voice qualities in infant duration experiments so as to not overstate the linear nature of duration increases.

Modal and non-modal voice quality classification using acoustic and electroglottographic features.

The goal of this study was to investigate the performance of different feature types for voice quality classification using multiple classifiers. The study compared the COVAREP feature set; which included glottal source features, frequency warped cepstrum and harmonic model features; against the mel-frequency cepstral coefficients (MFCCs) computed from the acoustic voice signal, acoustic-based glottal inverse filtered (GIF) waveform, and electroglottographic (EGG) waveform. Our hypothesis was that MFCCs can capture the perceived voice quality from either of these three voice signals. Experiments were carried out on recordings from 28 participants with normal vocal status who were prompted to sustain vowels with modal and non-modal voice qualities. Recordings were rated by an expert listener using the Consensus Auditory-Perceptual Evaluation of Voice (CAPE-V), and the ratings were transformed into a dichotomous label (presence or absence) for the prompted voice qualities of modal voice, breathiness, strain, and roughness. The classification was done using support vector machines, random forests, deep neural networks and Gaussian mixture model classifiers, which were built as speaker independent using a leave-one-speaker-out strategy. The best classification accuracy of 79.97% was achieved for the full COVAREP set. The harmonic model features were the best performing subset, with 78.47% accuracy, and the static+dynamic MFCCs scored at 74.52%. A closer analysis showed that MFCC and dynamic MFCC features were able to classify modal, breathy, and strained voice quality dimensions from the acoustic and GIF waveforms. Reduced classification performance was exhibited by the EGG waveform.

Characterisation of voice quality of Parkinson’s disease using differential phonological posterior features

Change in voice quality (VQ) is one of the first precursors of Parkinson’s disease (PD). Specifically, impacted phonation and articulation causes the patient to have a breathy, husky-semiwhisper and hoarse voice. A goal of this paper is to characterise a VQ spectrum – the composition of non-modal phonations – of voice in PD. The paper relates non-modal healthy phonations: breathy, creaky, tense, falsetto and harsh, with disordered phonation in PD. First, statistics are learned to differentiate the modal and non-modal phonations. Statistics are computed using phonological posteriors, the probabilities of phonological features inferred from the speech signal using a deep learning approach. Second, statistics of disordered speech are learned from PD speech data comprising 50 patients and 50 healthy controls. Third, Euclidean distance is used to calculate similarity of non-modal and disordered statistics, and the inverse of the distances is used to obtain the composition of non-modal phonation in PD. Thus, pathological voice quality is characterised using healthy non-modal voice quality “base/eigenspace”. The obtained results are interpreted as the voice of an average patient with PD and can be characterised by the voice quality spectrum composed of 30% breathy voice, 23% creaky voice, 20% tense voice, 15% falsetto voice and 12% harsh voice. In addition, the proposed features were applied for prediction of the dysarthria level according to the Frenchay assessment score related to the larynx, and significant improvement is obtained for reading speech task. The proposed characterisation of VQ might also be applied to other kinds of pathological speech.

Estimating Subglottal Pressure From Neck-Surface Acceleration During Normal Voice Production.

The purpose of this study was to evaluate the potential for estimating subglottal air pressure using a neck-surface accelerometer and to compare the accuracy of predicting subglottal air pressure relative to predicting acoustic sound pressure level (SPL). Indirect estimates of subglottal pressure (Psg') were obtained from 10 vocally healthy speakers during loud-to-soft repetitions of 3 different /p/-vowel gestures (/pa/, /pi/, /pu/) at 3 pitch levels in the modal register. Intraoral air pressure, neck-surface acceleration, and radiated acoustic pressure were recorded, and the root-mean-square amplitude of the acceleration signal was correlated with Psg' and SPL. The coefficient of determination between accelerometer level and Psg' was high when data were pooled from all vowel and pitch contexts for each participant (r2 = .68-.93). These relationships were stronger than corresponding relationships between accelerometer level and SPL (r2 = .46-.81). The average 95% prediction interval for estimating Psg' using accelerometer level was ±2.53 cm H2O, ranging from ±1.70 to ±3.74 cm H2O across participants. Accelerometer signal amplitude correlated more strongly with Psg' than with SPL. Future work is warranted to investigate the robustness of the relationship in nonmodal voice qualities, individuals with voice disorders, and accelerometer-based ambulatory monitoring of subglottal pressure.

Acquiring Complex Focus-Marking: Finnish 4- to 5-Year-Olds Use Prosody and Word Order in Interaction.

Using a language game to elicit short sentences in various information structural conditions, we found that Finnish 4- to 5-year-olds already exhibit a characteristic interaction between prosody and word order in marking information structure. Providing insights into the acquisition of this complex system of interactions, the production data showed interesting parallels to adult speakers of Finnish on the one hand and to children acquiring other languages on the other hand. Analyzing a total of 571 sentences produced by 16 children, we found that children rarely adjusted input word order, but did systematically avoid marked OVS order in contrastive object focus condition. Focus condition also significantly affected four prosodic parameters, f0, duration, pauses and voice quality. Differing slightly from effects displayed in adult Finnish speech, the children produced larger f0 ranges for words in contrastive focus and smaller ones for unfocused words, varied only the duration of object constituents to be longer in focus and shorter in unfocused condition, inserted more pauses before and after focused constituents and systematically modified their use of non-modal voice quality only in utterances with narrow focus. Crucially, these effects were modulated by word order. In contrast to comparable data from children acquiring Germanic languages, the present findings reflect the more central role of word order and of interactions between word order and prosody in marking information structure in Finnish. Thus, the study highlights the role of the target language in determining linguistic development.

Complex prosodic focus marking in Finnish: Expanding the data landscape

By investigating prosody beyond pitch and duration, this article provides a detailed and multifaceted picture of focus marking in a language that differs substantially from more extensively studied languages like English. A production study examined prosodic focus marking in Finnish based on acoustic analyses of 947 short SVO sentences spoken by 17 native speakers. The results indicated effects of information structure on five acoustic measures: f0, duration, intensity, the use of pauses and non-modal voice quality. Words in narrow focus had a larger f0 range, longer duration, larger intensity range and were followed by pauses more often than words in other information structural conditions. Conversely, contextually given words showed a smaller f0 range, shorter duration, and, in post-focal condition, lower intensity. Moreover, realisations with non-modal voice quality were more frequent for all syllables of post-focal given words compared to the broad focus condition, whereas for words in narrow focus, non-modal realisations were more frequent only on the last syllable. Observing these effects in parallel, the findings exceeded previous studies in scope, providing encouragement for a broader approach to the investigation of prosodic focus marking.

Voice quality in affect cueing: does loudness matter?

In emotional speech research, it has been suggested that loudness, along with other prosodic features, may be an important cue in communicating high activation affects. In earlier studies, we found different voice quality stimuli to be consistently associated with certain affective states. In these stimuli, as in typical human productions, the different voice qualities entailed differences in loudness. To examine the extent to which the loudness differences among these voice qualities might influence the affective coloring they impart, two experiments were conducted with the synthesized stimuli, in which loudness was systematically manipulated. Experiment 1 used stimuli with distinct voice quality features including intrinsic loudness variations and stimuli where voice quality (modal voice) was kept constant, but loudness was modified to match the non-modal qualities. If loudness is the principal determinant in affect cueing for different voice qualities, there should be little or no difference in the responses to the two sets of stimuli. In Experiment 2, the stimuli included distinct voice quality features but all had equal loudness to test the hypothesis that equalizing the perceived loudness of different voice quality stimuli will have relatively little impact on affective ratings. The results suggest that loudness variation on its own is relatively ineffective whereas variation in voice quality is essential to the expression of affect. In Experiment 1, stimuli incorporating distinct voice quality features consistently obtained higher ratings than the modal voice stimuli with varied loudness. In Experiment 2, non-modal voice quality stimuli proved potent in affect cueing even with loudness differences equalized. Although loudness per se does not seem to be the major determinant of perceived affect, it can contribute positively to affect cueing: when combined with a tense or modal voice quality, increased loudness can enhance signaling of high activation states.

Dialect experience in Vietnamese tone perception

This study investigated the perceptual dimensions of tone in Vietnamese and the effect of dialect experience on listener's prelinguistic perception of tone. While Northern Vietnamese tones are cued by a combination of pitch and voice quality, Southern Vietnamese tones are purely pitch based. 30 listeners from two Vietnamese dialects (10 Northern, 20 Southern) participated in a speeded AX discrimination task using northern stimuli. The resulting reaction times were used to compute an INDSCAL multidimensional scaling solution and were submitted to hierarchical clustering analysis. While the analysis revealed a similar three-dimensional perceptual space structure for both listener groups, corresponding roughly to f(0) offset, voice quality, and contour type, the relative salience of these dimensions varied by dialect: Southern listeners were more likely to confuse tones produced with nonmodal voice quality, whereas Northern listeners found tones with similar pitch excursions to be more confusable. The results of hierarchical clustering of the stimuli further support an analysis where low-level perceptual similarity is influenced by primary dialect experience.

Is nonmodal phonation at the end of utterances a cue for speaker recognition by humans?

Many studies of nonmodal phonation report a high rate of interspeaker variation. These experiments also showed that the rate and type of glottalization are characteristic to the speaker. Still, it remains an open question whether human listeners use this information as a cue for recognizing familiar voices. A listening experiment was conducted to investigate this issue, concentrating on utterance-final irregularities. Short utterances produced by four speakers were copy synthesized with the Klatt synthesizer in two conditions: with modal and with nonmodal voice quality in their final regions. Because two of the speakers were reliable ‘‘glottalizers’’ and two seldom ‘‘glottalized,’’ one of the two synthesis conditions reflected the speaker’s usual utterance-final voice quality and the other did not. These pairs of stimuli were used to test the effect of the presence or absence of utterance-final nonmodal phonation on the identification of voices familiar to the subjects. When asked to choose which one of the two variants is the speaker’s voice, listeners preferred the stimuli with the speaker’s typical voice quality at the end. Results will be discussed in comparison with a well-known perceptual cue on speaker identity, the F0 contour. [Work supported by a Fulbright Fellowship.]

Time Course of the First Formant Bandwidth

Phonetic characteristics of modal or non-modal voice qualities, such as open quotient, spectral tilt, the first formant bandwidth, and the degree of noise, have been measured in terms of the relative amplitude difference between the first two harmonics, the relative amplitude difference between the first harmonic and formant peaks, the first formant bandwidth, and harmonic-to-noise ratio, or subjective noise ratings. It was observed that one of the characteristics of breathy voice is an increased first formant bandwidth (Klatt & Klatt 1990) and that breathiness varies over time within a vowel (Silverman et al. 1995). It is predicted from these observations that the first formant bandwidth varies over time as well as across phonation types. This paper investigates the time course of the first formant bandwidth of the vowel /a/ in tokens of type /Ca/ and /NCa/, where C is 8 Korean coronal obstruents to see the behavior of the first formant bandwidth in those contexts.