Vocal folds vibratory patterns of “laryngeal mechanism M0” as investigated with high-speed cinematography and electroglottography

International audienceA human vocal production is characterized by the use of different laryngeal mechanisms(Roubeau,1993 ; Henrich et al., 2003), among which laryngeal mechanisms M1 and M2 arethe most frequently used and studied in speech and singing. Chest, or modal voice, isproduced in M1, while falsetto, or female head voice is produced in M2. Besides that, anothermechanism known as laryngeal mechanism 0 (M0) can also be found in speech and sometypes of singing. Vocal fry, pulse register, creaky voice, glottal fry are the terms frequentlyused to refer to this vocal production. Yet, these terms are sometimes considered assynonymous (Hollien, 1974), while in other cases, they would characterize different vocalproductions. The M0 mechanism has mainly been studied for pathological voices, but it isalso of interest for linguists, as it can be found in hesitations and at the end of utterances.Thus our aim is to investigate the different vibratory patterns that can be involved in M0thanks to high speed cinematography and to compare them with the EGG signal. The glottalvibratory movement has been simultaneously investigated in two ways: first, with the help ofelectroglottography (EGG), and secondly, with high-speed cinematography. The high-speedimages have been recorded using a high-speed camera unit Wolf HS Endocom 5560, and theEGG signal has been recorded using a Laryngograph Ltd. Two male speakers were asked toproduce a sustained vowel in M0 with or without vocal effort, and transitions between M1and M0. The high-speed images and the EGG signal were recorded simultaneously atdifferent sampling frequencies, respectively 4000 Hz and 44170 Hz. The data were thenprocessed using Matlab. The EGG signal was differentiated, and both EGG and DifferentiatedEGG (DEGG) signals were synchronized with the high-speed images, so as to compare theEGG and DEGG data with the vibratory movements observed on the images. An imageprocessing algorithm based on the detection of contrast (light energy threshold) wasdeveloped for relative measurement of glottal area, antero-posterior and transversal distances.It is a known fact that the glottal area function in M0 is characterized by short pulses, whichcan be single, double or triple within a fundamental period (Whitehead et al., 1984 ; Childersand Lee, 1991).On a physiological point of view, the thyroarytenoid muscles tend to shorten, subsequentlythickening the anterior-posterior dimension of the vocal folds, and allowing the anterior partof the vocal ligaments to vibrate (Hollien, 1974). However the physiological propertiesassociated with the multiple vibratory patterns mentioned above are not fully described,especially regarding the precise movements of the vocal folds margins, the degree of vocalfold adduction and the contribution of the ventricular folds.The analysis shows that different realizations of M0 can be found in this corpus especiallywhen comparing (i) the M0 realized with vocal effort from (ii) the M0 realized without vocaleffort. The first one is more likely to be characterized by a single pulse, while the second oneis frequently produced with multiple pulses - both realizations having different typical EGGshapes. We also found that in the case of single pulse pattern, the anterior portion of the vocalfolds is more involved in the vibrating process than the posterior one, while the posteriorportion of the vocal folds can also be active in the case of a multiple pulse pattern. Transitionsproduced by the two speakers were also investigated, thus revealing different which werepossibly used. The validity of EGG signals rendering the different vibratory patterns will bediscussed as well as implications for synthesis