Eur Arch Otorhinolaryngol (2007) 264:1201–1205

DOI 10.1007/s00405-007-0348-3

LARYNGOLOGY

Voice improvement in unilateral laryngeal paralysis during loud voicing: theoretical impact

Renaud Garrel · Richard Nicollas · Elodie Chapus · Maurice Ouaknine · Antoine Giovanni

Received: 17 September 2006 / Accepted: 4 May 2007 / Published online: 9 June 2007© Springer-Verlag 2007

Abstract Voice of patient with unilateral laryngeal paral-ysis (ULP) shows a nonlinear behaviour with suddenoctave jumps, bifurcations and chaos. Such a behaviourmay be due to an increased number of freedom degrees inthe glottal system. We hypothesized that voice intensity(with increasing sub glottal pressure) could improve vocalsignal stability with less freedom degrees in vibrating sys-tem, and then a decrease of nonlinearities. A prospectivestudy of 32 consecutive voices of patients with ULP andsevere dysphonia was conducted. Jitter and Lyapunovexponent from vocal signals were compared at comfortableand loud voicing with Wilcoxon’s test. In 23 out of 32patients, jitter signiWcantly decreased from 5 (median) innormal voice to 1.2 in loud voice (P < 10¡3), Lyapunovexponent decreased from 1,495 bit/s (median) to 708 bit/s(P < 10¡4). Two patients had paradoxical results regardingjitter (higher in loud voice) and 2 regarding Lyapunovexponent. From the 23 cases of voice improvement, 15cases showing a marked improvement of the acoustic anal-ysis supported our hypothesis (65%). Nonlinear phenomenadetected in vocal signals of ULP with severe dysphoniamay be reduced in loud voice.

Keywords Vocal fold vibration · Unilateral laryngeal paralysis · Coupling · Nonlinear dynamics · Lyapunov exponent

Introduction

Voice of patients with unilateral laryngeal paralysis (ULP)is characterized by association of excess of breath due toglottic leakage, and roughness, vibration being irregular[8]. But dysphonia varies and can be reduced by speechtherapy, endoscopic or surgical vocal fold medialization[11]. These methods reduce glottic leakage, insure bettercontact of vocal cords, and so a more regular signal.

Theories on phonation physiology explain this improve-ment by a better synchronization of vocal folds, but this isnot yet well known. Classic physical models, all more orless following Van den Berg [20], simply ignore the contactof vocal folds during the closed phase [18]. Since 1990,several authors have underlined nonlinear phenomena inpathologic vocal signal lacking in these models [19]. Inunilateral laryngeal paralysis, Herzel [12] showed suddenvoice breaks, octave jumps, and other nonlinear phenom-ena, signs of nonlinear functioning of larynx. All nonlinearphenomena express a discontinuous relation between phys-iological mechanisms for example and Wnal results.

For larynx, there are many nonlinear relations betweentension given to vocal fold and real tension achieved, sub-glottal pressure and transglottal Xow, collision phenomenabetween the vocal folds [16, 19]. Collision and interactionof vocal folds have been speciWcally studied by the IowaCity group in a work on provoked asymmetries in excisedanimal larynx. When applying diVerent tensions on thetwo vocal folds, there were diVerent stages in the function-ing of larynx, linked to the asymmetry and the subglottalpressure chosen [6]. When asymmetry increased, a highersubglottal pressure was required to produce a stable vibra-tion. So, threshold of phonatory pressure depends onasymmetry [17]. Just above the threshold, there remainedsome instability. Largely above it, the system was stable.

R. Garrel (&)Service ORL, CHU Gui de Chauliac, 34295 Montpellier, Francee-mail: r-garrel@chu-montpellier.fr

R. Nicollas · E. Chapus · M. Ouaknine · A. GiovanniFederation ORL, Laboratoire d’Audio Phonologie Clinique, UPRES-EA, CHU la Timone 264 r Saint Pierre, 13005 Marseille, France

123

1202 Eur Arch Otorhinolaryngol (2007) 264:1201–1205

So there clearly exists synchronization between the vocalfolds.

We wanted in our work to show how raised intensitycould improve voice in unilateral laryngeal paralysis, loudvoicing being linked to a decrease of nonlinear phenomenain the signal. So we followed Berry’s experimental plan, weanalysed vocal signal with classic index of vocal irregular-ity, the jitter [2, 14] but also with a “nonlinear” index,Lyapunov exponent we already found clinically pertinent[10]. In the present work, we hypothesized that voice inten-sity (with increasing sub glottal pressure) could improvevoice stability with less freedom degrees in vibrating sys-tem, and then a decrease of nonlinearities.

Patients and methods

We analysed the voice of 32 patients with ULP we saw inour Speech Therapy Department of La Timone with severedysphonias classiWed GRADE 3 (G scale of GRBAS sys-tem) with the usual criteria in our team: conversational talkduring consultation, evaluation by both doctor and speechtherapist always working as a team in the specializeddepartment.

All patients underwent videostroboscopy establishingULP. Aetiology was established following the guidelines ofFrench Society of Otolaryngology [11]. Electromyographywas not performed.

The population could be classiWed as follows:

– 20 women, 12 men.– Mean age 47 year (27–75).– Unknown aetiology: 11 (34%).– Post thoracic surgery: 16 (50%).– Post thyroid surgery: 5 (16%).

Perceptive and instrumental appreciation of the voice wasdone as usual in our group [21].

Patients were asked to send out three successively sus-tained vowels for at least 1 s

– “comfortable loud voice”,– “more loudly” and– “scream out”. (The latter is not studied here).

The voice was recorded with a Sennheiser® wired micro-phone, MD431 II (Old Lyme, CT, USA) in a noiselessroom, on DAT tape recorder (Tascam®DA-20 mk II,Tokyo, Japan). Voice samples were then digitised at a sam-pling frequency of 22 KHz with 32 bits amplitude withoutcompression in a.wav format by means of the Cool EditPro® software (Syntrillium™, Scottsdale, AZ, USA). Voicesamples were kept in a hard drive disk.

For vowels spoken at “comfortable loud voice” (condi-tion I), we chose the vowel we perceived nearest to the

patient’s ordinary voice. For the “more loudly” test, (condi-tion II), we chose the most loud and clear vowel giving bestincrease of the patient’s voice.

The mean fundamental frequency was F0 = 149 Hz incomfortable voice and raised to F0 = 164.2 Hz in loudvoice.

For both vowels, we had a perceptive analysis, followingthe G scale of GRBAS scale and performed by the samedoctor-speech therapist tandem seen at inclusion time in thesurvey. We measured jitter with the signal editor Atmos®

(Beta version of Atmos Medical™, Lenzkirch, Germany).We then measured Lyapunov exponent with speciWc soft-ware speciWcally used for this task in our lab [10]. Statisti-cal analysis allowed to compare results in condition I andII. For these small groups, we used the Wilcoxon matched-pairs signed-ranks test with p value threshold = 0.05.

Results

Perceptive analysis selected two groups of patients:

– A group A of 9 patients (28%) had no perceptive vocalimprovement when voice was louder, in condition II.

– On the other hand, group B with 23 (72%) had markedimprovement of voice: 17 patients from GRADE 3 toGRADE 2, 6 patients from GRADE 3 to GRADE 1.

Results of jitter analysis and Lyapunov exponent are onlygiven for the second group (Table 1). Jitter was obviouslyimproved from 5 (median) in condition I to 1.2 in conditionII, signiWcant diVerence with P = 0.000175 (Fig. 1). Lyapu-nov exponent was also improved, from 1,495 bit/s (median)to 708 bit/s, signiWcant diVerence with P = 0.000099 (Fig. 2).

Paradoxical results were only seen twice with jitterhigher in condition II, and also twice concerning Lyapunovexponent.

Discussion

We thought increasing voice intensity while improvingvoice quality could improve stability with less degrees offreedom in the vocal system. Out of our 32 patients withULP, 23 had a marked perceptual improvement of voicequality when voicing louder. If one considers that animprovement more than 20% is signiWcant for acousticalanalysis [2], our hypothesis is clearly supported in 15patients out of 23 (65%).

Assessment of vocal signal stability

Reliability of nonlinear analysis of vocal signals in dyspho-nia has been pointed out in the last ten years by authors

123

Eur Arch Otorhinolaryngol (2007) 264:1201–1205 1203

such as Titze et al. [19], Behrman and Baken [4, 5]. Ourteam displayed an algorithm to estimate Lyapunov expo-nent [10], witch is central in the work we now present. Wewill not here discuss “nonlinear” theories or methods

springing from chaos theories. Main comment about thisextensive research seems to be that perfectly symmetric andsynchronous functioning of vocal folds only exists in the-ory. Asymmetries, irregularities, “imperfections” of vocal

Table 1 Jitter and Lyapunov exponent values for “comfort-able loud” voice (condition I) and “loud” voice (condition II)

Patient numbers Jiitter Jitter Lyapunov Lyapunov Outliers

Condition I Condition II Condition I Condition II

1 6.5 0.8 1,914 758

2 6.5 1.25 1,070 330

3 6.2 0.71 2,200 2,025

4 5.8 1.2 2,150 708

5 5.7 2.1 2,028 328

6 5.5 2 1,970 430

7 5 0.95 1,714 709

8 3.8 1.6 1,370 695

9 3.1 0.49 980 810

10 1.8 0.91 1,836 609

11 1.7 1.25 1,948 1,501

12 0.85 0.82 820 1,138 a

13 3.6 0.93 1,280 1,138

14 3.8 1.76 1,320 640

15 2.5 1.2 580 730 a

16 2.2 0.89 501 350

17 4.5 1.23 1,500 900

18 4.3 0.85 1,490 750

19 4.5 0.73 1,495 985

20 5.7 0.67 2,020 570

21 3.4 1.34 1,210 660

22 1.23 2.45 2,120 308 a

23 1.01 2.01 530 528 a

Mean 3.88 1.22 1,480.26 765.22

Median 5.00 1.20 1,495.00 708.00

Standard deviation 1.83 0.53 542.46 399.30a Paradoxical results (outliers)

Fig. 1 Jitter variation for condition I and II. Outliers are in dashedlines

0

1

2

3

4

5

6

7

IInoitidnoCInoitidnoC

571000.0=p

rettij

Fig. 2 Lyapunov exponent variation (in bit/s) for condition I and II.Outliers are in dashed lines

0

005

0001

0051

0002

0052

IInoitidnoCInoitidnoC

ces/tib 990000.0=p

vonupayLtnenopxe

123

1204 Eur Arch Otorhinolaryngol (2007) 264:1201–1205

folds express as many degrees of freedom [4, 13]. Nonlin-ear method are designed to track in vocal signal the expres-sion of latent degrees of freedom and to give us new cluesto understand better physiopathology: fractal measures [7]and, on the other hand, Lyapunov exponent [10]. These areindicators of voice stability as jitter or shimmer, for exam-ple, but they do not need prior detection of F0, which variesso much in dysphonias [18]. One could note that the jitterfactor used here was F0 dependent but since the meanincrease of pitch was small (15.2 Hz), this phenomenonwas neglected.

Synchronization mechanisms

Louder voice in patients with ULP means better glottalfunction, increase of adduction and also aerodynamicforces, especially due to Bernoulli eVect. Interactionbetween the vocal folds may change the “trajectory” of thevibration. These nonlinear phenomena must be known.Coupling and synchronization of vocal folds are asserted bya decrease in jitter and Lyapunov exponent. ULP seems agood clinical model to study the coupling of the vocalfolds. We especially wanted to understand how voice couldget better despite the diVerence in vocal folds muscular ten-sion. Asymmetry of tension and vibrating mass betweennormal and pathologic vocal fold tend to desynchronisationbut coupling factors, if weakened, still remain relativelystrong. In ULP, voice is known to be acceptable if theimpaired vocal fold is close enough to the other, which isachieved during adequate manipulation by a speech thera-pist or after surgical medialization. Coupling then voicequality might be improved if vocal folds adduction wasstronger in spite of the asymmetric tension. This maydecrease the number of freedom degrees in the system. Weasked our patients with ULP to increase voice intensity. Wesuppose they can do so by increasing glottal closure andsubglottic pressure, contracting thyroarytenoid muscle, asshown in physiologic laryngeal vibration [18]. We studiedthis on excised larynx [9]. Larynx with asymmetric tensionsends an acoustic signal where subharmonics shown byspectral analysis may tell of a nonlinear combinationbetween the vocal folds. A new experimental study withoptoreXectometer strengthens that idea of synchronizationof the two vocal folds [15]. When tension is asymmetric,each vocal fold moves with a diVerent vibration speed dur-ing the open phase of the cycle. At closure time, the twosignals tend to synchronise. Actually, we imagined thesephenomena may exist in every larynx, none showing per-fect symmetry in mass or tension, just as perfect symmetryis never reached for both hands or both halves of the face[18]. Physical contact of the vocal folds is the best knowncoupling factor. The two folds share some elasticity andmass characteristics, contact is enhanced by surface mucus,

its viscosity being an important factor [1]. But BernoullieVect is also a major cause of synchronization, nearly sym-metric on both vocal folds because of their geometry [18].

Usually, glottis can be seen as a set of coupled oscilla-tors. If coupling is perfect, the system is stable withoutnoise, irregularity or chaos. It becomes impossible to trackthe signal. But when asymmetry is important for tension(ULP) or for mass (polyps or other expansive lesions) cou-pling is less eVective, the signal becomes what in physics iscalled a “nonlinear combination”, the system, potentiallyunstable, exhibits sensitive dependence on initial conditionsas systems belonging to Welds of deterministic chaos [13].Such a system displays a higher number of freedom degrees[22].

Theoretical impact

Our study conWrms clinically David Berry’s initial experi-ments [6]. When vocal folds are asymmetric, vibration,near the threshold, is unstable and irregular with higher jit-ter and Lyapunov exponent. That is the zone of spontane-ous voice for patients with ULP and severe dysphonia.Beyond the threshold, vibration can become stable despiteasymmetric tension of the vocal folds, if higher couplingforces make up for the asymmetry. We studied variableintensity and asymmetry in the interaction of the vocalfolds. When vibrations are unstable, vibrating laryngealsystem is a chaos with a high number of freedom degrees[22]. For Baken, this unstable vibration reXects hyperdi-mensionality [3]. Ultra rapid video would now be useful toshow interaction and coupling in the vibrating physiologyof larynx. Theoretical models did not really study coupling;larynx was seen as two coexisting hemilaynxes, with onlyvirtual contact, and no inXuence of a vocal fold on theother. Actually, these models describe undulation of themucosal membrane during the open phase, but not closureand contact of the two folds exchanging energy. Investiga-tion should be resumed to try and explain nonlinear cou-pling phenomena, function of glottis vibration, and energyexchange between subglottic pressure and vibrating part ofthe vocal fold.

References

1. Ayache S, Ouaknine M, Dejonkere P, Prindere P, Giovanni A(2004) Experimental study of the eVects of surface mucus viscos-ity on the glottic cycle. J Voice 18:107–115

2. Baken RJ, OrlikoV RF (1999) Clinical measurement of speech andvoice, 2nd edn, p 624

3. Baken RJ, OrlikoV RF (2003) Curing diagnosis: improving thetaxonomy of phonatory dysfunction. In: 6th conference on advanc-es in quantitative laryngology April 2003. Hamburg, Germany

4. Behrman A (1999) Global and local dimensions of vocal dynam-ics. J Acoust Soc Am 105:432–443

123

Eur Arch Otorhinolaryngol (2007) 264:1201–1205 1205

5. Behrman A, Baken RJ (1997) Correlation dimension of electro-glottographic data from healthy and pathologic subjects. J AcoustSoc Am 102:2371–2379

6. Berry DA, Herzel H, Titze IR, Story BH (1996) Bifurcations in ex-cised larynx experiments. J Voice 10:129–138

7. Boek W, Wieneke GH, Dejonckere PH (1997) Clinical relevanceof the fractal dimension of F0 perturbations computed by the box-counting method. J Voice 11:437–442

8. Dejonckere PH, Bradley P, Clemente P et al (2001) A basic proto-col for functional assessment of voice pathology, especially forinvestigating the eYcacy of (phonosurgical) treatments and evalu-ating new assessment techniques. Guideline elaborated by theCommittee on Phoniatrics of the European Laryngological Society(ELS). Eur Arch Otorhinolaryngol 258:77–82

9. Giovanni A, Ouaknine M, Guelfucci R, Yu T, Zanaret M, TrigliaJM (1999) Nonlinear behavior of vocal fold vibration: the role ofcoupling between the vocal folds. J Voice 13:465–476

10. Giovanni A, Ouaknine M, Triglia JM (1999) Determination oflargest Lyapunov exponents of vocal signal: application to unilat-eral laryngeal paralysis. J Voice 13:41–54

11. Guerrier B (2003) Pathologie de la corde vocale de l’adulte. Rap-port de la Société Française d’Oto-Rhino-laryngologie et Chirur-gie Cervico-Faciale ed Arnette, Paris

12. Herzel H (1993) Bifurcations and chaos in voice signals. ApplMech Rev 46:399–413

13. Herzel H, Berry D, Titze I, Steinecke I (1995) Nonlinear dynamicsof the voice: signal analysis and biomechanical modeling. Chaos5:30–34

14. Kent RD (2003) The MIT Encyclopedia of Communication Disor-ders, Cambridge

15. Ouaknine M, Garrel R, Giovanni A (2003) Separate detection ofvocal fold vibration by optoreXectometry: a study of biphonationon excised porcine larynges. Folia Phoniatr Logop 55:28–38

16. SataloV RT (2005) Voice science. Plural Publishing Ed San Diego,CA, p 308

17. Titze IR (1992) Phonation threshold pressure: a missing link inglottal aerodynamics. J Acoust Soc Am 91:2926–2935

18. Titze IR (1994) Principles of voice production. Prentice–Hall,Englewood CliVs

19. Titze IR, Baken RJ, Herzel H (1993) Evidence of chaos in vocalfold vibration. In: Principle of voice production. Frontiers in basicscience San Diego, CA, pp 143–188

20. Van den Berg JW (1958) Myoelastic-aerodynamic theory of voiceproduction. J Speech Hear Res 1:227–244

21. Yu P, Revis J, Wuyts FL, Zanaret M, Giovanni A (2002) Correla-tion of instrumental voice evaluation with perceptual voice analy-sis using a modiWed visual analog scale. Folia Phoniatr Logop54:271–281

22. Zhang Y, Jiang JJ (2005) Spatiotemporal chaos in excised larynxvibrations. Phys Rev E Stat Nonlin Soft Matter Phys 72:035201

123