Marek Frič 1

1 Acadamy of Performing Arts of Prague, Musical acoustic research center, Prague

A gain factor (GF) is determined as an amount of the spectral band increasing related to overall sound pressure level. It has been published dependence of GF on the vocal loudness, subglottal pressure or vocal effort variation, acoustic spaces, gender and age differences. All that studies analyzed LTAS of a long time samples in the same voice tasks (reading, singing or shouting) at voice loudness variation.

Our study examines changes of GF in 3 different voice tasks (habitual speaking voice, supported speaking voice and singing) and in 3 groups of voice users, separately for both genders: professional operatic singers (3 male, 6 females), professional actors (7 m., 4 f.) and as a reference nonprofessional group of 20 students of 1^st class of acting (11 m. and 9 f.).

GF was calculated from third-octave spectral characteristics of manually separated long vowels [a:] (approximately 300 ms long) from gradually increasing of loudness of the word [ma:ma] pronounce in constant pitch. The GF was calculated for all pitches from frequency range of the subject and accepted was only with linear regression coefficient of p<0.05 significance level. Than for the subject was calculated mean overall GF in a given spectral band from all pitches GFs. Mean overall GF values was compared by Student’s test for 3 tasks and 3 groups. As relevant band differences was chosen only that with p<0.05 level.

Generally there has been found no statistical difference between habitual and supported speaking GF in all groups of subjects, but there has occurred evident difference between speaking and singing. In singing, the students (males and females) had higher GFs of lower (0.5 – 1.5) spectral bands; professional male actors had higher GFs in 0.6 and 2-3 kHz bands, but actresses in 1-2 kHz. Operatic singers have referred only differences between supported speaking and singing, when males had higher GFs in 0.6, 0.95 and 5 kHz bands and females in 0.6-0.8 and 2-3 kHz bands.

The comparison of different groups reveals main difference between singers and actors.

In habitual voice, male professional singers increased more the 1.5 – 2 kHz spectral bands than male actors (professionals and students). In opposite professional females (actresses and singers) have had higher GF of the spectral bands over the 3 kHz than female students.

In supported voice, professional actors gained more: 0.75 and 1.5 kHz males, females only 0.75 kHz band; operatic male singers had higher GFs in 1.5 – 2 kHz but lower in 0.75 and 3 kHz; female singers higher in 3 and over 7 kHz bands than males and females students.

In singing, professional male (female) operatic singers had higher values of GF in all spectral bands over 3 kHz than male (female) students, but actors have higher GFs in 1.2 and 3 kHz and actress only in 0.6 kHz band.

Results show that different groups use different spectral gain strategies in different voice tasks.

Weikert Matthias 1 , Schlömicher-Thier Josef 2 , Andrea Orendi 1

1 GesundheitsForum Regensburg, HNO-Phoniatrician-group praxis, Regensburg

2 Austrian Voice Institute e.v., occupational Voice, Salzburg/ Regensburg

On the basis of voice range profile measurement (VPR) with the PC-Program of Lingwaves we demonstrate the approved methode of VPR . VPR, acoustic and aerodynamic measurements and the Laryngo-stroboscopy are the main parameters of acoustic data in the ELS-evaluation of voice diagnostic (LRO 2005/84). Concerning to ELS the ENT specialist in voice-diagnostics focused has to measure the  professional voice user´s  individual capacity of seaking-, singing, and shouting voice, including DSI, with the aim of optimal advising, counselling and treatment. We present a case report study of 71 students and aspirants of speech communication, and music-pedagogy of the University of Regensburg, Department of Rhetorik und Sprecherziehung in the year of 2007-2008. VPR is an important module of the five parameteres in ELS- evaluation of voice. The importance and the relevance of VPR /DSI for singers and speakers controversely are discussed. Finally the well trained ear and  the performing arts experience of the ENT –doctor / phoniatrician is the main goal to advise and treat professional voice users like the consultants of speech and communication.

Wivine Decoster 1 , Piet Mertens 2

1 K.U.Leuven, Expertisecentrum Stem, Leuven

2 K.U.Leuven, Linguistics, Leuven

As prosody comprises a variety of components (pitch, loudness, duration, pauses), which contribute to intonation, stress, tempo and rhythm, it plays a central role in the liveliness of speech and in expressing contrasts according to the speaker's personality, linguistic meaning and structure, expressive tension and communicative intention.

Problems with prosody (dysprosody) are a substantial part of many speech language pathologies. Therapy interventions mostly aim at elaborating the vocabulary, correcting the syntax, refining the pronunciation, extending communicative skills, developing coping strategies.  However prosody has a key function in speech intelligibility as it decreases when prosody deteriorates and remains good as long as prosody is adequate, even with disordered speech. On the one hand dysprosody can be caused by or be the cause of speech language pathology (SLP) problems. On the other hand, prosody can be a facilitator during SLP-therapy.

A crucial concept in experimental and clinical prosody research is the measuring method. Usually SLP diagnosis protocols are based on auditory perceptual evaluation with a variety of variables, scales and categories. This results in a qualitative overall impression that reflects the holistic nature of auditory perception in daily life. However, for research and diagnosis guided intervention, quantitative data are needed. Instrumental prosody measures have rarely been the focus of scientific concern in SLP and voice science. Where acoustic data are involved, the relevance for the perception of prosody remains unclear.  

To tackle these problems, while meeting the high standards of scientific research, one may benefit from the information provided by the Prosogram, a software tool developed by P. Mertens (Linguistics, K.U.Leuven). The tool provides a semi-automatic transcription of prosody, based on a stylization of the fundamental frequency data (contour) for the syllabic nuclei in the speech signal. The stylization simulates tonal perception of human listeners (Mertens, P., 2004; Mertens, P., 2004b). The Prosogram is implemented as a script for PRAAT, a free open software package for the acoustic analysis of speech (Boersma, P., 2001).

During PEVOC8 we will present the Prosogram: its methodology (F0-extraction, segmentation, graphical presentation), its reliability and validity, its use and its benefits for voice research.

Ahmed Geneid 1 , Marjo Rönkkö 1 , Paavo Alku 2 , Risto Voutilainen 3 , Liisa Airaksinen 3 , Erkki Vilkman 1

1 Department of Phoniatrics, Helsinki University Hospital, Phoniatrics Department, Helsinki

2 Finnish Institute of Occupational Health, , Helsinki

3 Helsinki University of Technology, Department of Signal Processing and Acoustics, Espoo

The aim of the research was to study if there are acoustic findings for inaudible changes in the symptoms of the vocal organ. This was done through analyzing 180 voice samples from nine subjects (five females) recorded before and after exposure to a placebo substance (Lactose) and an organic dust substance. From a previous research by the authors, subjects reported subjective changes in their voices upon exposure to organic dust substances which were not detected through perceptual assessment by voice clinicians. Such symptoms included feeling that voice is hoarse,  husky, tense or feeling extra effort when speaking and difficulty to start phonation (p<0.05). Other significant symptoms included feeling of shortness of breath or need to gasp for air and feeling that voice is weak or that it does not resonate (p<0.01). The important finding in the research resulted from acoustical analysis that was done for the voice samples using inverse filtering. Results showed that the values of parameter OQ1 changes significantly after exposure to the organic dust substance (p<0.01). Such change draws the attention to possible changes in the vocal organ that maintains the end product of the vocal organ through changes that can not be perceived acoustically.

Keywords: Organic dust - voice - acoustical analysis - perceptual assessment - inverse filtering - vocal organ

Peter Pabon 1 , Sten Ternström 2

1 Royal Conservatory, Inst. of Sonology, The Hague

2 Royal Institute of Technology, KTH, Dept of Speech, Music and Hearing, TMH, Stockholm

According to Titze (1992), the source-filter model predicts that the maximum contour of the voice range profile (VRP) should exhibit intensity ripples that are due to harmonic-formant interactions (HFI). Gramming and Sundberg (1988) initially suggested the phenomenon to be responsible for local peaks and troughs they observed in the upper VRP contour. This prompted Klingholz (1988) to question the importance of HFI. He reclaimed that the main upper contour rippling instead would be due to the register transition between modal and falsetto voice. The debate then went dormant. From singing practice, it is clear that optimal HFI will increase the total power of the sung vowel, but in VRP recording practice this is rarely seen. Except for a dip, directly associated with a change in register, the upper contour of the VRP typically does not ripple consistently. In manual VRP recording, the sampling in fundamental frequency is usually too sparse for such an effect to be observed. 

Might HFI rippling of the upper VRP contour in fact be common, without being recognized as such, due to its great variability? 60 automatic VRP recordings of both trained and untrained voices were stored with additional narrow band spectrum information for each semitone interval. This spectral information was used to partition the upper VRP contour into frequency subranges, each with one specific dominant harmonic. Averaged upper VRP contour curves are presented that were first aligned according to their dominant harmonic. These curves show the importance and the average size of HFI induced ripples in the upper VRP contour, and provide additional insight as to whether or not singers and non-singers tune formants to partials. 

Jan G. Svec 1 , Svante Granqvist 2

1 Palacky University, Dept.Exp.Physics, Biophysics, Olomouc, Czech Republic

2 Royal Institute of Technology, Speech, Music and Hearing, Stockholm, Sweden