What kind of phonation causes the strongest vocal fold collision? – A hemi-larynx phonation contact pressure study

2.1 Sensor and read out unit

A commercial pressure mapping sensor (4201, Tekscan Inc., South Boston, MA, USA) with a piezoresistive sensing area of 21.1 × 45.7 mm² and a density of 27.6 sensels cm⁻² is chosen to measure the occurring pressure during phonation [18]. The manufacturer specifies the measurable pressure range up to 34 kPa. In order to fulfil the needs for the application, a custom read out unit was developed and a circuit board was designed [19].

The read out unit can be seen in Figure 1. It is based on two multiplexers switching between a reduced number of sensels, so the measured area was reduced to 15.2 mm × 30.4 mm and the sample rate could be increased. The setup is described in detail by Scheible et al. [19]. It was calibrated to pressures up to 17.6 kPa with a sample rate up to 1.2 kHz.

2.2 Anatomy & physiology of the larynx and vocal folds

The VFs are located inside the larynx and have a length of 10 mm–20 mm in anterior-posterior direction and a width of 8 mm–12 mm in medial-lateral direction [20]. The air from the lungs which flows through the VFs, forces them to vibrate. This vibration, also known as phonation, produces a pressure signal which is the sound source of the human voice. The main frequency of the vibration is called the fundamental frequency f₀. To change the sound, this vibration can be manipulated with laryngeal muscles.

In the human speech different types of phonation exist, Sundberg [13] distinguished between them with the help of the glottal resistance, where a low glottal resistance indicates a “breathy” and a high glottal resistance a “pressed” voice. Further a “breathy” voice is indicated by the failure of the VFs to make contact. Therefore a “breathy” phonation will not be measured in this investigation. So when the denominations are used it should be understood as a phonation trending towards one of those types.

2.3 Larynx preparation

At the local slaughterhouse eleven larynges were gathered and immediately quick-frozen with 2-methylbutane (−150 °C). Shortly before the measurement they were unfrozen at 4 °C and later put in a NaCl solution for 15 min at room temperature. Superfluous tissue and the vestibular folds were removed to enable a straight view on the VF during phonation. In order to place the contact plate with the sensor along the original glottal midline the left VF was removed. The preparation was done according to Döllinger et al. [21].

The plate and the larynx were placed in the holder and mounted by screws to secure its position. The remaining holes and gaps between larynx and plate, where the air from the sub-glottal space could escape, is closed with protefix, an adhesive paste for dentures prosthesis.

Figures 2 and 6 show the larynx mounted in the holder. If the vestibular folds are removed, the quality of the phonation suffers due to the lost weight of the removed tissue. In previous phonation studies a stabilization weight of 10 g is added and could produce good results [22].

Figure 1:

Pressure sensing matrix and the custom made read out unit.

Figure 2:

Snapshot of the highspeed camera. The marker points can be seen. On the right side the pressure sensor matrix is illustrated.

2.4 Image recognition

Subsequently to the larynx preparation the VF is tattooed with marker points. In Figure 2 the inked markers as well as the measured length l = P 1 P 5 ̄ and width w = P 3 P 4 ̄ are indicated. The positions of the marker points are analyzed manually by selecting the marker position using a custom software.

2.5 Experimental setup

The setup is similar to the one described in a publication of our group where the sensor was already integrated in a hemi-larynx phonation experiment [19]. The prepared larynx is positioned in the holder which sits over the artificial trachea, where the remaining part of the porcine trachea is connected to. On the side of the removed VF the sensor, mounted on a plate, was positioned with a slight contact to the remaining VF. This can be seen in Figure 2. A mass flow controller then regulates the flow rate through the larynx. In order to prevent the biological tissue from drying out a humidifier was installed. The onset flow rate was found by increasing the flow until a clear sustained phonation was audible.

During the measurement series with four equidistant flow steps the sub-glottal pressure, the audio signal (96 kS/s) as well as the contact pressure (1.2 kS/s) were recorded with National Instruments-measurement cards. The movement of the VF was recorded with a high-speed camera (4 kS/s) from above. All flow rates were maintained for 5 s. After a settle-time the recording was started for 0.5 s. The setup is shown in Figure 3.

Figure 3:

The experimental setup: the hemi-larynx is mounted on the artificial trachea inside the holder. The medial surface of the VF is oscillating against the pressure-sensing matrix mounted on a planar surface. From above, the high-speed camera records the movement and the microphone the sound of the vibrating VF.

2.6 Manipulations

During speech, the VFs undergo continuous manipulations. Important manipulations are adduction and elongation [11]. In a living species, these manipulations are performed by laryngeal muscles. An adduction is managed by an outer rotation of the arytenoid cartilage forced by the lateral cricoarytenoid muscle. An elongation is caused by a forward rotation of the thyroid cartilage caused by the contraction of the cricothyroid muscle [11]. To mimic these muscle activities, threads have been stitched in the arytenoid cartilage and thyroid cartilage. On the end of those threads weights apply force in order to adduct or elongate the VF, similar as it was done in previous studies [22].

Every larynx is manipulated in the same order, see Figure 4. It starts with a reference measurement Ref1, followed by adductions with 30 g (A30) and 50 g weights (A50). Then a second reference measurement Ref2 is made, before the larynx is elongated with 30 g (E30), 50 g (E50) and 70 g (E70). Lastly a third reference measurement Ref3 is made, this enables an observation of a potential hysteresis. All of those reference measurements were made with 10 g adduction.

Figure 4:

Cyclic order of manipulations.

2.7 Data

3.1 Exemplary data of a single measurement

The contact pressures of two sensels (compare Figure 6), the sub-glottal pressure and the glottal amplitude are plotted in Figure 5. The moment of a complete glottis closure is indicated by a sudden pressure drop followed by a quick rise as can be seen in the sub-glottal pressure data. These instances are illustrated with dashed lines. All frequency domain data show peaks at 128 Hz, where the sub-glottal pressure and the glottal amplitude also shows higher spectral components, which are multiples of 128 Hz. These data were already shown in a former publication by our group [19]. Regarding the contact pressure wave forms, it can be seen, that a first plateau is followed by a high contact peak, subsequently the VF opens again. Similar curves can be found in literature gathered by measurements but also simulation [11, 15, 26].

Figure 5:

In the upper plot, the contact pressure of two sensels are drawn over a time range of 0.1 s, the middle one shows the subglottal pressure and in the lower one the glottal amplitude is plotted. Black dashed vertical lines mark the closing of the VF which is indicated by a sudden pressure drop followed by a quick rise. As one can observe it is also the moment of the contact pressure peak. On the right side the frequency domain of the whole measurement duration of 0.5 s are shown.

Figure 6:

Two sensels of the eighth row, one in the fifth column (5/8) and next to it in the sixth (6/8) are selected for further demonstration. The eighth row is the bottom most and is hidden in behind the VF.

3.2 Statistical analysis

Out of eleven prepared larynges only seven delivered data for all measurements and could be used for evaluation. The others did not show a clear phonation or a phonation was not possible within all manipulations.

The larynges did have a different onset flowrate, due to their size differences, the mean value of the onset flow rate was 15.4 slm (standard liter per minute).

3.2.1 Deformation of the VF

The position of the marker points gives insight how the manipulation effects the VF. In Figure 7 the change of the tissue normalized to the initial measurement can be seen. For the adduction it seems that the VF gets 3.0 % shorter but the lateral surface strain increases about 6.0 %. The following elongation thins the VFs width about 3.2 % and length increases due to the stretching of 6.0 % of the VF.

Figure 7:

This boxplot shows the change VF of the length in blue and the width in green. The measurements are normalized on the first reference. The boxes indicate the 25th and 75th percentiles, whiskers show the most extreme data points that are included. The horizontal bar in the box indicates the median, excluded outliers are marked with a red ‘+’.

3.2.2 Effect of manipulations on the fundamental frequency

In Figure 8 the fundamental frequencies can be seen. The frequency starts initially for the first reference measurement Ref1 at a median value of 258.9 Hz. The manipulations do change the frequency. It can be stated, that the obtained adduction levels do show a higher effect on the frequency and do also lead to higher “frequencies” 366.9 Hz compared to the applied elongation where the highest frequency was measured to 300.3 Hz at an elongation with 50 g. The frequency then falls to 200.8 Hz for the last reference measurement.

Figure 8:

Effect of manipulations on the fundamental frequency f₀.

3.2.3 Glottal resistance

Analogous to the frequency the glottal resistance rises stronger for the adduction. In Figure 9 can be seen that it starts with a median of 123.0 kPa slm⁻¹ to increase subsequently to a maximum of 143.5 kPa slm⁻¹ with an adduction of 50 g. Applying the elongation the glottal resistance stays around the value of the second reference measurement of 113.6 kPa slm⁻¹ and finally drops to 86.2 kPa slm⁻¹ for the last unmanipulated reference measurement.

Figure 9:

Boxplot of the glottal resistance for each manipulation.

3.2.4 Closing velocity

In Figure 10 the velocity of the VF for each manipulation can be seen. The values do show a strong trend for different flow rates, but do not react so strongly due to the manipulations. To demonstrate this, the flow rates have been illustrated separately. But one can see that the glottal velocity shows lower values for the adduction and then slightly higher values for the elongation.

Figure 10:

Boxplot of the max. closing velocity for different manipulations, grouped by the flow steps.

3.2.5 Contact pressures

The contact pressure at 50 % of the glottal midline is evaluated and the maximum values are plotted for all manipulations, flow rates and larynges in Figure 11. It can be seen, that with some exceptions, a higher flow rate results in a higher contact pressure in general. The highest median value of 0.34 kPa is reached with an elongation of 50 g. Further it is to notice that the last reference value does not fall back to the initial values of the two other reference measurements.

Figure 11:

Boxplot of the maximum contact pressures for different manipulations, grouped by the flow steps.

3.3 Correlation between different measured parameters

A data set of seven larynges, with eight manipulations and four flowsteps could be collected, which results in 224 single measurements. This made a statistical evaluation to find correlations underneath different parameters possible. A Pearson-correlation was applied on the data set. As one can see in Figure 12, the contact pressure correlates weakly (0.3 < r < 0.5) with the closing velocity as well as with the flow rate, and it shows a negative correlation with the glottal resistance (r = −0.29). Furthermore it can be seen that the closing velocity correlates strongly with the flowrate (r > 0.61). The extensive experimental setup delivers even more parameters; only the most significant are chosen for demonstrations and those that do not show correlations, were not plotted.

Figure 12:

The four parameters contact pressure, closing velocity, glottal resistance and flow rate are plotted against each other and a Pearson-correlation among each other is investigated. The p value to neglect the null hypothesis was set to p = 0.05.