Academia.eduAcademia.edu

Outline

Title
Abstract
Figures
The Analysis and Synthesis Processes
Overview of the Analysis-Synthesis Process
Limitations and Conclusions
Supplemental Materials
References

Integrated software for analysis and synthesis of voice quality

Profile image of Bruce GerrattBruce Gerratt
https://doi.org/10.3758/BRM.42.4.1030
visibility

…

description

12 pages

link

1 file

Abstract

Voice quality is an important perceptual cue in many disciplines, but knowledge of its nature is limited by a poor understanding of the relevant psychoacoustics. This article (aimed at researchers studying voice, speech, and vocal behavior) describes the UCLA voice synthesizer, software for voice analysis and synthesis designed to test hypotheses about the relationship between acoustic parameters and voice quality perception. The synthesizer provides experimenters with a useful tool for creating and modeling voice signals. In particular, it offers an integrated approach to voice analysis and synthesis and allows easy, precise, spectral-domain manipulations of the harmonic voice source. The synthesizer operates in near real time, using a parsimonious set of acoustic parameters for the voice source and vocal tract that a user can modify to accurately copy the quality of most normal and pathological voices. The software, user's manual, and audio files may be downloaded from http://brm.psychonomic-journals.org/content/supplemental. Future updates may be downloaded from www.surgery.medsch.ucla.edu/glottalaffairs/.

Figures (7)
Figure 2. The synthesizer interface. (A) The work window. (B) Detail of the toolbar. Time-domain source manipulations are accom- plished by clicking and dragging the different Liljencrants—Fant parameters, which are plotted with dots on the flow derivative pulses in the middle right frame in the synthesizer window (labeled “A” in Figure 2A). “B” shows the vocal tract model, and “C” shows the superimposed spectra of the natural and synthetic voices. Adjustments can be made to vocal tract resonances and bandwidths, the noise-to-signal ratio, jitter level, and/or shimmer level, either by sliding the appropriate cursor or by typing the desired value into the appropriate cell in the table, labeled “D.” See the text for more discussion.
Figure 2. The synthesizer interface. (A) The work window. (B) Detail of the toolbar. Time-domain source manipulations are accom- plished by clicking and dragging the different Liljencrants—Fant parameters, which are plotted with dots on the flow derivative pulses in the middle right frame in the synthesizer window (labeled “A” in Figure 2A). “B” shows the vocal tract model, and “C” shows the superimposed spectra of the natural and synthetic voices. Adjustments can be made to vocal tract resonances and bandwidths, the noise-to-signal ratio, jitter level, and/or shimmer level, either by sliding the appropriate cursor or by typing the desired value into the appropriate cell in the table, labeled “D.” See the text for more discussion.
Figure 3. The Liljencrants—Fant (LF) model fitting process. The spectrum of the LF-fitted pulse is calculated and displayed (show: as “A”) along with the raw output of the inverse filter and the smooth, LF-fitted pulse (“B”). Jitter (in percent) is modeled using a shape-preserving interpolation algorithm that expands or contracts the glot- Traditionally, two sources of spectral noise have been distinguished. The first is noise related to irregularities in vocal-fold vibration (jitter and shimmer, representing random variability in the period and amplitude of glottal pulses, respectively; see, e.g., Baken, 1987, for a review). Noise also emerges due to turbulence generated during the open phase of the glottal cycle and/or flow through a per- sistent glottal gap (especially for female and pathological voices). Noise is often measured separately from jitter and
Figure 3. The Liljencrants—Fant (LF) model fitting process. The spectrum of the LF-fitted pulse is calculated and displayed (show: as “A”) along with the raw output of the inverse filter and the smooth, LF-fitted pulse (“B”). Jitter (in percent) is modeled using a shape-preserving interpolation algorithm that expands or contracts the glot- Traditionally, two sources of spectral noise have been distinguished. The first is noise related to irregularities in vocal-fold vibration (jitter and shimmer, representing random variability in the period and amplitude of glottal pulses, respectively; see, e.g., Baken, 1987, for a review). Noise also emerges due to turbulence generated during the open phase of the glottal cycle and/or flow through a per- sistent glottal gap (especially for female and pathological voices). Noise is often measured separately from jitter and
Figure 4. Modeling /0 (“A”) and amplitude (“B”) contours. The top frames show the unsmoothed output of the pitch and amplitude trackers, the middle frames show the smoothed pitch and amplitude tracks, and the bottom frame shows the residuals. After a synthetic voice has been created (or a case opened), the operator is able to adjust all parameters to modify the voice as desired. These modifications can serve a number of purposes, including creation of stimuli to test listeners’ sensitivity to different acoustic param- eters and their assessment of the accuracy and validity of acoustic measurement techniques. Some of these applica- tions are described later in the present article. Synthesizing the voice. Because the synthesizer sam- pling rate is currently fixed at 10 kHz, the following pro- cedure is applied to overcome quantization limits on mod- eling f0. First, a plot of f0 versus time is generated for the duration of the 1-sec token to be synthesized, taking into account the desired pitch contour, tremor, jitter, and shim- mer. For each successive glottal pulse, f0 is determined by interpolating this curve at the starting instant for that
Figure 4. Modeling /0 (“A”) and amplitude (“B”) contours. The top frames show the unsmoothed output of the pitch and amplitude trackers, the middle frames show the smoothed pitch and amplitude tracks, and the bottom frame shows the residuals. After a synthetic voice has been created (or a case opened), the operator is able to adjust all parameters to modify the voice as desired. These modifications can serve a number of purposes, including creation of stimuli to test listeners’ sensitivity to different acoustic param- eters and their assessment of the accuracy and validity of acoustic measurement techniques. Some of these applica- tions are described later in the present article. Synthesizing the voice. Because the synthesizer sam- pling rate is currently fixed at 10 kHz, the following pro- cedure is applied to overcome quantization limits on mod- eling f0. First, a plot of f0 versus time is generated for the duration of the 1-sec token to be synthesized, taking into account the desired pitch contour, tremor, jitter, and shim- mer. For each successive glottal pulse, f0 is determined by interpolating this curve at the starting instant for that
Figure 5. Manipulating the voice source in the spectral domain. In this example, the first two harmonics have been selected, as indi. cated by the line segment in panel A, and the amplitude of the first harmonic has been increased, as shown by the arrow. The increas in harmonic amplitude is indicated by a second arrow in panel B. Panel C shows the derivative of the glottal pulse corresponding t¢ the spectrum in A. The slope of the line segment (in this case, 9.409 dB) appears in the cell labeled “Source Spectral Slope” in the up: permost right corner of the synthesis window. Figure 5 shows the procedure for manipulating the voice source in the spectral domain. The amplitude or slope of an individual harmonic or a group of harmonics is manipulated by first selecting the desired harmonics, which are then connected by a line segment (indicated by an arrow in panel A of Figure 5). Users then click and drag the line segment to raise or lower either endpoint, thus changing the slope of that group of harmonics. The resulting increase in the amplitude of the first this example is indicated by an arrow in panel of the line segment (in dB) will appear in the corner of the syn Alternatively, it is possible to change the slo ment to obtainas harmonic in B. The slope upper right hesis window (labeled “C” in Figure 5). pe of a seg- pecific target value by typing the desired value into cell C and then clicking the line segment. Users can also raise or the absolute amp ower the segment as a who while preserving e to change itudes of all of the selected harmonics, heir relative amplitudes. The same pro- cedures apply when manipulating any number of harmon- ics or the entire spectral slope at once. Any number of seg- ihe synthesizer allows the source to be manipulated in both the time and spectral domains. Time-domain ma- nipulations are accomplished by clicking and dragging the different LF parameters, which are plotted with dots on the flow-derivative pulses in the top right frame in the syn- thesizer window (labeled “A” in Figure 2A). However, it is often desirable to edit the harmonic voice source in the spectral domain as well as (or instead of) in the time do- main. Although theoretical correspondences between LF model parameters and details of source spectral shape have been investigated (Fant, 1995; Ni Chasaide & Gobl, 1997), in practice, it is quite difficult to manipulate combinations of LF parameters in the time domain to achieve a specific desired spectral change. Further, although the LF model can accommodate a wide variety of source functions, it is not so flexible that it can model the complete range of glottal pulse shapes that occur even among normal speak- ers (Henrich et al., 2001; Shue, Kreiman, & Alwan, 2009). For these reasons, the synthesizer permits the source to be modified in the frequency (spectral) domain as well as in the time domain. This increased flexibility with respect to source shapes greatly improves accuracy and ease of
Figure 5. Manipulating the voice source in the spectral domain. In this example, the first two harmonics have been selected, as indi. cated by the line segment in panel A, and the amplitude of the first harmonic has been increased, as shown by the arrow. The increas in harmonic amplitude is indicated by a second arrow in panel B. Panel C shows the derivative of the glottal pulse corresponding t¢ the spectrum in A. The slope of the line segment (in this case, 9.409 dB) appears in the cell labeled “Source Spectral Slope” in the up: permost right corner of the synthesis window. Figure 5 shows the procedure for manipulating the voice source in the spectral domain. The amplitude or slope of an individual harmonic or a group of harmonics is manipulated by first selecting the desired harmonics, which are then connected by a line segment (indicated by an arrow in panel A of Figure 5). Users then click and drag the line segment to raise or lower either endpoint, thus changing the slope of that group of harmonics. The resulting increase in the amplitude of the first this example is indicated by an arrow in panel of the line segment (in dB) will appear in the corner of the syn Alternatively, it is possible to change the slo ment to obtainas harmonic in B. The slope upper right hesis window (labeled “C” in Figure 5). pe of a seg- pecific target value by typing the desired value into cell C and then clicking the line segment. Users can also raise or the absolute amp ower the segment as a who while preserving e to change itudes of all of the selected harmonics, heir relative amplitudes. The same pro- cedures apply when manipulating any number of harmon- ics or the entire spectral slope at once. Any number of seg- ihe synthesizer allows the source to be manipulated in both the time and spectral domains. Time-domain ma- nipulations are accomplished by clicking and dragging the different LF parameters, which are plotted with dots on the flow-derivative pulses in the top right frame in the syn- thesizer window (labeled “A” in Figure 2A). However, it is often desirable to edit the harmonic voice source in the spectral domain as well as (or instead of) in the time do- main. Although theoretical correspondences between LF model parameters and details of source spectral shape have been investigated (Fant, 1995; Ni Chasaide & Gobl, 1997), in practice, it is quite difficult to manipulate combinations of LF parameters in the time domain to achieve a specific desired spectral change. Further, although the LF model can accommodate a wide variety of source functions, it is not so flexible that it can model the complete range of glottal pulse shapes that occur even among normal speak- ers (Henrich et al., 2001; Shue, Kreiman, & Alwan, 2009). For these reasons, the synthesizer permits the source to be modified in the frequency (spectral) domain as well as in the time domain. This increased flexibility with respect to source shapes greatly improves accuracy and ease of
Note—All parameters except those indicated have been held constant across all stimuli. Synthetic Voice Stimuli Created by Varying Individual Synthesizer Parameters Table 1
Note—All parameters except those indicated have been held constant across all stimuli. Synthetic Voice Stimuli Created by Varying Individual Synthesizer Parameters Table 1

Related papers

Reproducibility of Voice Parameters: The Effect of Room Acoustics and Microphones
Eric Hunter

Journal of Voice, 2018

Introduction. Computer analysis of voice recordings is an integral part of the evaluation and management of voice disorders. In many practices, voice samples are taken in rooms that are not sound attenuated and/or soundproofed ; further, the technology used is rarely consistent. This will likely affect the recordings, and therefore, their analyses. Objectives. The objective of this study is to compare various acoustic outcome measures taken from samples recorded in a soundproofed booth to those recorded in more common clinic environments. Further, the effects from six different commonly used microphones will be compared. Methods. Thirty-six speakers were recorded while reading a text and producing sustained vowels in a controlled acoustic environment. The collected samples were reproduced by a Head and Torso Simulator and recorded in three clinical rooms and in a sound booth using six different microphones. Newer measures (eg, Pitch Strength, cepstral peak prominence, Acoustic Voice Quality Index), as well as more traditional measures (eg Jitter, Shimmer, harmonics-to-noise ratio and Spectrum Tilt), were calculated from the samples collected with each microphone and within each room. Results. The measures which are more robust to room acoustic differences, background noise, and microphone quality include Jitter and smooth cepstral peak prominence, followed by Shimmer, Acoustic Voice Quality Index, harmonics-to-noise ratio, Pitch Strength, and Spectrum Tilt. Conclusions. The effect of room acoustics and background noise on voice parameters appears to be stronger than the type of microphone used for the recording. Consequently, an appropriate acoustical clinical space may be more important than the quality of the microphone.

View PDFchevron_right
The multidimensional nature of pathologic voice quality.
Bruce Gerratt

Although the terms "breathy" and "rough" are frequently applied to pathological voices, widely accepted definitions are not available and the relationship between these qualities is not understood. To investigate these matters, expert listeners judged the dissimilarity of pathological voices with respect to breathiness and roughness. A second group of listeners rated the voices on unidimensional scales for the same qualities. Multidimensional scaling analyses suggested that breathiness and roughness are related, multidimensional constructs. Unidimensional ratings of both breathiness and roughness were necessary to describe patterns of similarity with respect to either quality. Listeners differed in the relative importance given to different aspects of voice quality, particularly when judging roughness. The presence of roughness in a voice did not appear to influence raters' judgments of breathiness; however, judgments of roughness were heavily influenced by the degree of breathiness, the particular nature of the influence varying from listener to listener. Differences in how listeners focus their attention on the different aspects of multidimensional perceptual qualities apparently are a significant source of interrater unreliability (noise) in voice quality ratings.

View PDFchevron_right
Perception of synthesized voice quality in connected speech by Cantonese speakers
Kathryn M Hird

The Journal of the Acoustical Society of America, 2002

Perceptual voice analysis is a subjective process. However, despite reports of varying degrees of intrajudge and interjudge reliability, it is widely used in clinical voice evaluation. One of the ways to improve the reliability of this procedure is to provide judges with signals as external standards so that comparison can be made in relation to these ''anchor'' signals. The present study used a Klatt speech synthesizer to create a set of speech signals with varying degree of three different voice qualities based on a Cantonese sentence. The primary objective of the study was to determine whether different abnormal voice qualities could be synthesized using the ''built-in'' synthesis parameters using a perceptual study. The second objective was to determine the relationship between acoustic characteristics of the synthesized signals and perceptual judgment. Twenty Cantonese-speaking speech pathologists with at least three years of clinical experience in perceptual voice evaluation were asked to undertake two tasks. The first was to decide whether the voice quality of the synthesized signals was normal or not. The second was to decide whether the abnormal signals should be described as rough, breathy, or vocal fry. The results showed that signals generated with a small degree of aspiration noise were perceived as breathiness while signals with a small degree of flutter or double pulsing were perceived as roughness. When the flutter or double pulsing increased further, tremor and vocal fry, rather than roughness, were perceived. Furthermore, the amount of aspiration noise, flutter, or double pulsing required for male voice stimuli was different from that required for the female voice stimuli with a similar level of perceptual breathiness and roughness. These findings showed that changes in perceived vocal quality could be achieved by systematic modifications of synthesis parameters. This opens up the possibility of using synthesized voice signals as external standards or ''anchors'' to improve the reliability of clinical perceptual voice evaluation.

View PDFchevron_right
Analysis by synthesis of severely pathological voices
Bruce Gerratt

The ability to synthesize pathological voices may provide a tool for the development of a standard protocol for assessment of vocal quality. An analysis-by-synthesis approach using the Klatt formant synthesizer was applied to study 24 tokens of the vowel /a/ spoken by males and females with moderate-to-severe voice disorders. Both temporal and spectral features of the natural waveforms were analyzed and the results were used to guide synthesis. Perceptual evaluation indicated that about half the synthetic voices matched the natural waveforms they modeled in quality. The stimuli that received poor ratings reflected failures to model very unsteady or “gargled” voices or failures in synthesizing perfect copies of the natural spectra. Several modifications to the Klatt synthesizer may improve synthesis of pathological voices. These modifications include providing jitter and shimmer parameters; updating synthesis parameters as a function of period, rather than absolute time; modeling diplophonia with independent parameters for fundamental frequency and amplitude variations; providing a parameter to increase low-frequency energy; and adding more pole-zero pairs.

View PDFchevron_right
The perceptual structure of pathological voice quality
Bruce Gerratt

Although perceptual assessment is included in most protocols for evaluating pathologic voices, a standard set of valid scales for measuringvoice quality has never been established. Standardization is important for theory and for clinical acceptance, and also because validation of objective measures of voice depends on valid perceptual measures. The present study used large sets (n=80) of male and female voices, representing a broad range of diagnoses and vocal severities. Eight experts judged the dissimilarity of each pair of voices, and responses wereanalyzed using nonmetric individual differences multidimensional scaling. Results indicate that differences between listeners in perceptualstrategy are so great that the fundamental assumption of a common perceptual space must be questioned. Because standardization depends on the assumption that listeners are similar, it is concluded that effortsto standardize perceptual labels for voice quality are unlikely to succeed. However, analysis by synthesis may provide an alternate means of modeling quality as a function of both voices and listeners, thus avoiding this problem. (C) 1996 Acoustical Society of America.

View PDFchevron_right
The multidimensional nature of pathologic vocal quality
Gerald Berke

The Journal of the Acoustical Society of America, 1994

Although the terms "breathy" and "rough" are frequently applied to pathological voices, widely accepted definitions are not available and the relationship between these qualities is not understood. To investigate these matters, expert listeners judged the dissimilarity of pathological voices with respect to breathiness and roughness. A second group of listeners rated the voices on unidimensional scales for the same qualities. Multidimensional scaling analyses suggested that breathiness and roughness are related, multidimensional constructs. Unidimensional ratings of both breathiness and roughness were necessary to describe patterns of similarity with respect to either quality. Listeners differed in the relative importance given to different aspects of voice quality, particularly when judging roughness. The presence of roughness in a voice did not appear to influence raters' judgments of breathiness; however, judgments of roughness were heavily influenced by the degree of breathiness, the particular nature of the influence varying from listener to listener. Differences in how listeners focus their attention on the different aspects of multidimensional perceptual qualities apparently are a significant source of interrater unreliability (noise) in voice quality ratings.

View PDFchevron_right
Analysis by synthesis of pathological voices using the Klatt synthesizer
Abeer Alwan

Speech Communication, 1997

The ability to synthesize pathological voices may provide a tool for the development of a standard protocol for assessment of vocal quality. An analysis-by-synthesis approach using the Klatt formant synthesizer was applied to study 24 tokens of the vowel rar spoken by males and females with moderate-to-severe voice disorders. Both temporal and spectral features of the natural waveforms were analyzed and the results were used to guide synthesis. Perceptual evaluation indicated that about half the synthetic voices matched the natural waveforms they modeled in quality. The stimuli that received poor ratings reflected failures to model very unsteady or ''gargled'' voices or failures in synthesizing perfect copies of the natural spectra. Several modifications to the Klatt synthesizer may improve synthesis of pathological voices. These modifications include providing jitter and shimmer parameters; updating synthesis parameters as a function of period, rather than absolute time; modeling diplophonia with independent parameters for fundamental frequency and amplitude variations; providing a parameter to increase low-frequency energy; and adding more pole-zero pairs. q 1997 Elsevier Science B.V.

View PDFchevron_right
Transformation of voice quality in singing using glottal source features
Alexsandro Meireles

2019

Glottal activity information can be very important in several speech processing applications, such as in speech therapy , voice disorder diagnosis, voice transformation and text-to-speech synthesis. However, the use of algorithms for estimating glottal parameters from the speech signal is very limited in those applications because of problems with robustness and accuracy. In singing synthesis, the glottal source representation is also very important because it is closely related with the emotions and singing style. This paper proposes a robust method to estimate the voice quality parameters of the glottal source by using both the electroglottographic signal and the acoustic recordings of singing voice for five vowels in three different voice qualities: modal, breathy and creaky. The analysis of the resulting measurements permitted to confirm that voice quality parameters of the glottal source are correlated with the type of voice. Moreover, another experiment was conducted to show that it is possible to transform the modal singing voice into breathy and creaky by using an analysis-synthesis method that incorporates a glottal source model.

View PDFchevron_right
Dept. for Speech, Music and Hearing Quarterly Progress and Status Report Waweform and spectrum of the glottal voice source
Johan Sundberg

2000

The singer's formant has been found to be an important acoustic characteristic of professional operatic altos, tenors, baritones and basses, while it seems unclear whether or not sopranos possess it. According to Dmitriev & Kiselev (1 979) the JFequencies of the "high singing formant" increases @om lower voices to higher voices. By means of a listening test with synthesised vowel stimuli we examined how important the center JFequency of the singer's formant is to voice classijication. The results showed that it clearly influenced voice classijication. In another listening test with synthesised vowel stimuli we examined if a singer's formant belongs to the typical characteristics of a soprano voice. The results showed that increasing the center JFequency of the singer's formant up to 3.4 kHz increased the perceived quality of the soprano synthesis. For the stimulus containing the highest pitches, a synthesis lacking a singer's formant was preferred, while for other stimuli the results were less clear. The frequency relationships between the lower formants and the spectrum partials, especially for theJirst and last tone of a scale, also seem to be relevant for the quality rating of the soprano synthesis.

View PDFchevron_right
Source model adequacy for pathological voice synthesis
Bruce Gerratt

Pathological voices are particularly difficult to inverse filter and fit with source models, in part, because of source-tract interactions inherent in these voice types. In order to obtain good synthetic copies of these voices, we need to know the practical importance of accurately inverse filtering and modeling individual voice sources. To this end, thirty different voices were synthesized, each with four different sources. Preliminary results suggest that the output of the inverse filter does not always contain enough information about the source to adequately reconstruct vocal quality. However, the LF model of the voice source pulse does provide enough degrees of freedom to model naturally-occurring quality variations across voices.

View PDFchevron_right

References (50)

  1. for the glottal source waveform. Proceedings of the IEEE International Conference on Acoustical Speech Signal Processing, 1605-1608.
  2. Gerratt, B. R., & Kreiman, J. (2001). Measuring vocal quality with speech synthesis. Journal of the Acoustical Society of America, 110, 2560-2566. doi:10.1121/1.1409969
  3. Gerratt, B. R., Till, J., Rosenbek, J. C., Wertz, R. T., & Boysen, A. E. (1991). Use and perceived value of perceptual and instrumental measures in dysarthria management. In C. A. Moore, K. M. Yorkston, & D. R. Beukelman (Eds.), Dysarthria and apraxia of speech: Per- spectives on management (pp. 77-93). Baltimore: Brookes.
  4. Goldinger, S. D., Pisoni, D. B., & Logan, J. S. (1991). On the nature of talker variability effects on recall of spoken word lists. Journal of Experimental Psychology: Learning, Memory, & Cognition, 17, 152- 162. doi:10.1037/0278-7393.17.1.152
  5. Gordon, M., & Ladefoged, P. (2001). Phonation types: A cross- linguistic overview. Journal of Phonetics, 29, 383-406.
  6. Hanson, H. M. (1997). Glottal characteristics of female speakers: Acoustic correlates. Journal of the Acoustical Society of America, 101, 466-481. doi:10.1121/1.417991
  7. Hawks, J. W., & Miller, J. D. (1995). A formant bandwidth estimation procedure for vowel synthesis. Journal of the Acoustical Society of America, 97, 1343-1344.
  8. Henrich, N., D'Alessandro, C., & Doval, B. (2001). Spectral cor- relates of voice open quotient and glottal flow asymmetry: Theory, limits and experimental data. EUROSPEECH 2001, 47-51.
  9. Huffman, M. K. (1987). Measures of phonation type in Hmong. Jour- nal of the Acoustical Society of America, 81, 495-504.
  10. Iseli, M., & Alwan, A. (2004). An improved correction formula for the estimation of harmonic magnitudes and its application to open quotient estimation. In Acoustics, Speech, and Signal Processing (ICASSP 2004 Proceedings), pp. 669-672.
  11. Javkin, H. R., Antoñanzas-Barroso, N., & Maddieson, I. (1987). Digital inverse filtering for linguistic research. Journal of Speech & Hearing Research, 30, 122-129.
  12. Karlsson, I. (1991). Female voices in speech synthesis. Journal of Pho- netics, 19, 111-120.
  13. Kawahara, H. (1997). Speech representation and transformation using adaptive interpolation of weighted spectrum: Vocoder revisited. Pro- ceedings of the IEEE International Conference on Acoustics, Speech, & Signal Processing, 2, 1303-1306.
  14. Kempster, G. B., Gerratt, B. R., Verdolini Abbott, K., Barkmeier- Kraemer, J., & Hillman, R. (2009). Consensus auditory-perceptual evaluation of voice: Development of a standardized clinical proto- col. American Journal of Speech-Language Pathology, 18, 124-132. doi:10.1044/1058-0360(2008/08-0017)
  15. Klatt, D. H. (1980). Software for a cascade/parallel formant synthe- sizer. Journal of the Acoustical Society of America, 67, 971-975.
  16. Klatt, D. H., & Klatt, L. C. (1990). Analysis, synthesis, and percep- tion of voice quality variations among female and male talkers. Jour- nal of the Acoustical Society of America, 87, 820-857. doi:10.1121/ 1.398894
  17. Kreiman, J., & Gerratt, B. R. (2005). Perception of aperiodicity in pathological voice. Journal of the Acoustical Society of America, 117, 2201-2211.
  18. Kreiman, J., & Gerratt, B. R. (2010). Perceptual sensitivity to first harmonic amplitude in the voice source. Journal of the Acoustical Society of America, 128, 2085-2089.
  19. Kreiman, J., Gerratt, B. R., Iseli, M., Neubauer, J., Shue, Y.-L., & Alwan, A. (2008, August). The relationship between open quotient and H * 1 -H * 2 . In Proceedings of the 6th International Conference on Voice Physiology & Biomechanics. Tampere, Finland.
  20. Kreiman, J., Gerratt, B. R., Kempster, G. B., Erman, A., & Berke, G. S. (1993). Perceptual evaluation of voice quality: Review, tutorial, and a framework for future research. Journal of Speech & Hearing Research, 36, 21-40.
  21. Kreiman, J., Gerratt, B. R., & Khan, S. D. (2010). Effects of native language on perception of voice quality. Journal of Phonetics, 38, 588-593. doi:10.1016/j.wocn.2010.08.004
  22. Krishnan, A., & Gandour, J. T. (2009). The role of the auditory brain- stem in processing linguistically relevant pitch patterns. Brain & Lan- guage, 110, 135-148. doi:10.1016/j.bandl.2009.03.005
  23. Michaelis, D., Gramss, T., & Strube, H. W. (1997). Glottal-to-noise REFERENCES
  24. Airas, M. (2008). TKK Aparat: An environment for voice inverse fil- tering and parameterization. Logopedics Phoniatrics Vocology, 33, 49-64. doi:10.1080/14015430701855333
  25. Ananthapadmanabha, T. V. (1984). Acoustic analysis of voice source dynamics. Speech Transmission Laboratory Quarterly Progress & Status Report, 25(2-3), 1-24.
  26. Andruski, J., & Ratliff, M. (2000). Phonation types in production of phonological tone: The case of Green Mong. Journal of the Interna- tional Phonetic Association, 30, 37-61.
  27. Baken, R. J. (1987). Clinical measurement of speech and voice. Boston: College Hill.
  28. Bangayan, P., Long, C., Alwan, A. A., Kreiman, J., & Gerratt, B. R. (1997). Analysis by synthesis of pathological voices using the Klatt synthesizer. Speech Communication, 22, 343-368. doi:10.1016/ S0167-6393(97)00032-0
  29. Belotel-Grenié, A., & Grenié, M. (2004). The creaky voice phonation and the organisation of Chinese discourse. International Symposium on Tonal Aspects of Languages: With Emphasis on Tone Languages (ISCA, Beijing, China), pp. 5-8.
  30. Colton, R. H., Casper, J. K., & Leonard, R. (2005). Understanding voice problems: A physiological perspective for diagnosis and treat- ment. Baltimore: Lippincott Williams & Wilkins.
  31. Cutler, A., Dahan, D., & van Donselaar, W. (1997). Prosody in the comprehension of spoken language: A literature review. Language & Speech, 40, 141-201.
  32. de Krom, G. (1993). A cepstrum-based technique for determining a harmonics-to-noise ratio in speech signals. Journal of Speech & Hear- ing Research, 36, 254-266.
  33. Fant, G. (1960). Acoustic theory of speech production. The Hague: Mouton.
  34. Fant, G. (1979). Glottal source and excitation analysis. Speech Trans- mission Laboratory Quarterly Progress & Status Report, 20(1), 85- 107.
  35. Fant, G. (1995). The LF-model revisited. Transformations and fre- quency domain analysis. Speech Transmission Laboratory Quarterly Progress & Status Report, 36(2-3), 119-156.
  36. Fant, G., Liljencrants, J., & Lin, Q. (1985). A four-parameter model of glottal flow. Speech Transmission Laboratory Quarterly Progress & Status Report, 26(4), 1-13.
  37. Fischer-Jörgensen, E. (1967). Phonetic analysis of breathy (mur- mured) vowels in Gujarati. Indian Linguistics, 28, 71-139.
  38. Fitch, W. T., Neubauer, J., & Herzel, H. (2002). Calls out of chaos: The adaptive significance of nonlinear phenomena in mammalian vocal production. Animal Behaviour, 63, 407-418. doi:10.1006/ anbe.2001.1912
  39. Fujisaki, H., & Ljungqvist, M. (1986). Proposal and evaluation of excitation ratio: A new measure for describing pathological voices. Acustica, 83, 700-706.
  40. Ní Chasaide, A., & Gobl, C. (1997). Voice source variation. In W. J. Hardcastle & J. Laver (Eds.), The handbook of phonetic sciences (pp. 427-461). Oxford: Blackwell.
  41. Nygaard, L. C., & Pisoni, D. B. (1998). Talker-specific learning in speech perception. Perception & Psychophysics, 60, 355-376.
  42. Qi, Y., & Hillman, R. E. (1997). Temporal and spectral estimations of harmonics-to-noise ratio in human voice signals. Journal of the Acoustical Society of America, 102, 537-543.
  43. Rothenberg, M. (1973). A new inverse-filtering technique for deriving the glottal air flow waveform during voicing. Journal of the Acoustical Society of America, 53, 1632-1645. doi:10.1121/1.1913513
  44. Rothenberg, M. (1977). Measurement of airflow in speech. Journal of Speech & Hearing Research, 20, 155-176.
  45. Shrivastav, R. (2003). The use of an auditory model in predicting per- ceptual ratings of breathy voice quality. Journal of Voice, 17, 502-512. doi:10.1067/S0892-1997(03)00077-8
  46. Shue, Y.-L., Kreiman, J., & Alwan, A. (2009). A novel codebook search technique for estimating the open quotient. Interspeech 2009, 2895-2898.
  47. Stevens, K. N. (1998). Acoustic phonetics. Cambridge, MA: MIT Press.
  48. Strik, H. (1998). Automatic parameterization of differentiated glottal flow: Comparing methods by means of synthetic flow pulses. Journal of the Acoustical Society of America, 103, 2659-2669.
  49. Strik, H., & Boves, L. (1992). On the relation between voice source parameters and prosodic features in connected speech. Speech Com- munication, 11, 167-174.
  50. Volodina, E. V., Volodin, I. A., Isaeva, I. V., & Unck, C. (2006). Biphonation may function to enhance individual recognition in the dhole, Cuon alpinus II. Ethology, 112, 815-825. doi:10.1111/j.1439 -0310.2006.01231.x

Related papers

Perceptual evaluation of voice quality: review, tutorial, and a framework for future research
Bruce Gerratt

The reliability of listeners' ratings of voice quality is a central issue in voice research because of the clinical primacy of such ratings and because they are the standard against which other measures are evaluated. However, an extensive literature review indicates that both intrarater and interrater reliability fluctuate greatly from study to study. Further, our own data indicate that ratings of vocal roughness vary widely across individual clinicians, with a single voice often receiving nearly the full range of possible ratings. No model or theoretical framework currently exists to explain these variations, although such a model might guide development of efficient, valid, and standardized clinical protocols for voice evaluation. We propose a theoretical framework that attributes variability in ratings to several sources (including listeners' backgrounds and biases, the task used to gather ratings, interactions between listeners and tasks, and random error). This framework may guide development of new clinical voice and speech evaluation protocols, ultimately leading to more reliable perceptual ratings and a better understanding of the perceptual qualities of pathological voices.

View PDFchevron_right
Measuring voice quality with speech synthesis
Bruce Gerratt

Much previous research has demonstrated that listeners do not agree well when using traditional rating scales to measure pathological voice quality. Although these findings may indicate that listeners are inherently unable to agree in their perception of such complex auditory stimuli, another explanation implicates the particular measurement method-rating scale judgments-as the culprit. An alternative method of assessing quality-listener-mediated analysis-synthesis-was devised to assess this possibility. In this new approach, listeners explicitly compare synthetic and natural voice samples, and adjust speech synthesizer parameters to create auditory matches to voice stimuli. This method is designed to replace unstable internal standards for qualities like breathiness and roughness with externally presented stimuli, to overcome major hypothetical sources of disagreement in rating scale judgments. In a preliminary test of the reliability of this method, listeners were asked to adjust the signal-to-noise ratio for 12 synthetic pathological voices so that the resulting stimuli matched the natural target voices as well as possible For comparison to the synthesis judgments, listeners also judged the noisiness of the natural stimuli in a separate task using a traditional visual-analog rating scale. For 9 of the 12 voices, agreement among listeners was significantly (and substantially) greater for the synthesis task than for the rating scale task. Response variances for the two tasks did not differ for the remaining three voices. However, a second experiment showed that the synthesis settings that listeners selected for these three voices were within a difference limen, and therefore observed differences were perceptually insignificant. These results indicate that listeners can in fact agree in their perceptual assessments of voice quality, and that analysis-synthesis can measure perception reliably.

View PDFchevron_right
Defining and measuring voice quality
Bruce Gerratt

Although voices provide listeners with significant information about speakers, defining and quantifying voice quality remain elusive goals. The ANSI standard definition of quality (that attribute of auditory sensation in terms of which a listener can judge that two sounds similarly presented and having the same loudness and pitch are dissimilar) is often criticized because it specifies what quality is not, rather than what it is. It has also proven difficult to devise measurement protocols for quality as specified in the ANSI definition. We argue that the ANSI definition is in fact appropriate, because it treats quality as the result of perceptual processesinteractions between listeners and signals in the context of specific perceptual goals. Application of speech synthesis in method-of-adjustment tasks allows measurement of quality psychoacoustically as those aspects of the signal that allow a listener to determine that two sounds of equal pitch and loudness are different, consistent with the ANSI definition, and provides insight into the salient acoustic attributes contributing to quality. This technique holds promise for improving the reliability and validity of measures of voice quality. WHAT IS VOICE? THE DEFINITIONAL DILEMMA The speaking voice naturally conveys information about the speaking individual, and voice quality serves as a primary means by which speakers project their physical, psychological, and social characteristics to the world (Laver, 1980). Although voice is central to a range of human activities, it has proven difficult to provide a single, useful, all-purpose definition of voice, and several senses of the term are in common use. Voice can be defined narrowly as sound produced by vibration of the vocal folds, or broadly as essentially synonymous with speech. Details of phonation and articulation, pitch and amplitude variations, and temporal patterning all contribute to how a speaker sounds, and broad definitions of voice reflect this fact. Which stage in the voice production process receives definitional focus depends on the interest of the practitioner, experimenter, or listener.

View PDFchevron_right
Perceptual Evaluation of Voice Quality
Gerald Berke

Journal of Speech, Language, and Hearing Research, 1993

The reliability of listeners’ ratings of voice quality is a central issue in voice research because of the clinical primacy of such ratings and because they are the standard against which other measures are evaluated. However, an extensive literature review indicates that both intrarater and interrater reliability fluctuate greatly from study to study. Further, our own data indicate that ratings of vocal roughness vary widely across individual clinicians, with a single voice often receiving nearly the full range of possible ratings. No model or theoretical framework currently exists to explain these variations, although such a model might guide development of efficient, valid, and standardized clinical protocols for voice evaluation. We propose a theoretical framework that attributes variability in ratings to several sources (including listeners’ backgrounds and biases, the task used to gather ratings, interactions between listeners and tasks, and random error). This framework may g...

View PDFchevron_right
Perceptual assessment of voice quality: Past, present, and future
Bruce Gerratt, Jody Kreiman

Despite many years of research, we still do not know how to measure vocal quality. This paper reviews the history of quality assessment, describes some reasons why current approaches are unlikely to be fruitful, and proposes an alternative approach that addresses the primary difficulties with existing protocols.

View PDFchevron_right
580 California St., Suite 400
San Francisco, CA, 94104
© 2025 Academia. All rights reserved