II Electroglottography

Electroglottography (EGG) is a technique used to register laryngeal behavior indirectly by a measuring the change in electrical impedance across the throat during speaking. The method was first developed by Fabre (1957) and influential contributions are credited to Fourcin (1971 with Abberton) and Frokjaer-Jensen (1968 with Thorvaldsen). Commercially available devices are produced by Laryngograph Ltd., Synchrovoice and F-J Electronics.

8. How the device operates

A high frequency electrical current of small voltage and amperage (physiologically safe) passes between two electrodes situated on the surface of the throat at the level of the alae of the thyroid cartilage. The basic configuration of the device is depicted in Fig.14.

Figure 14. The principle of the EGG device (from Childers & Krishnamurthy, 1985:133).

The electrodes are made of copper, silver or gold. They have the form of rings or rectangles covering an area ranging from 3 cm2 to 9 cm2. A third electrode is often used as a reference for impedance measurements. It may be designed as a separate electrode or as a ring electrode encircling each of the two other electrodes. The electrodes are usually mounted on a flexible band whose length may be adjusted to hold the electrodes in a steady position and to still allow the subject to comfortably speak and breathe naturally. Sometimes the electrodes are mounted on a small holder which is pressed against the throat by hand. A signal generator supplies the electrodes with an AC sinusoidal current of an alternating frequency usually ranging from 300 kHz to 5 MHz. This frequency is sufficiently high, so that the current capacitancively bypasses the less conductive skin layer without the use of additional conductive paste (Rothenberg & Mahshie, 1988). The generator may produce constant voltage or constitute a constant current source (Childers & Krishnamurthy, 1985:132). The supplied current is different for each particular device, but is not stronger than several milliamperes. The voltage between the electrodes depends on the tissue impedance but the typical value is about 0.5 V (ibid.). In accordance with Fourcin, Hirose (in: Hardcastle & Laver, 1995:124) states that a power dissipation of only several microwatts occurs at the level the subject's vocal folds.

The sensing electrode detects the current as it passes through the skin and the throat. The percentage of amplitude modulation of the received signal reflects the percentage change in tissue impedance in the current's path (ibid.). The received signal is then demodulated by a signal detector circuit. The typical signal-to-noise ratio of the demodulator is about 40 dB. The demodulated waveform is then A/D converted and stored in a computer. Suitable additions to the standard configuration of the device consists of instruments for measuring the signal strength (e.g. a light-emitting diode) or for the measuring of signal symmetricity, showing the relation between the signals emitted by the two electrodes (Synchrovoice and F-J Electronics devices). This is very convenient for the proper placing of the electrodes.

The rapid variation in the conductance is caused mainly by the movement of the vocal folds. As they are separated, the transversal electrical impedance is high due to the fact that air impedance is much higher than tissue impedance. As they approximate and the contact between them increases, the impedance decreases, which results in a relatively higher current flow through the larynx structures. At the maximum contact the decrease is about 1% (up to 2%) of the total larynx conductance (Baken, 1992). According to Childers and Krishnamurthy (1985) the reason for the current modulation effect is a longer tissue passage for the radio frequency current when the glottis is open, since the total impedance of the tissue is a function of the length of the tissue passage:"We generally believe the impedance is least for full fold contact because under this condition there are, in effect, many parallel equally conductive resistance paths between the electrodes. The combined total parallel resistance is less than the resistance of any one path. Therefore, it is reasonable to postulate that the tissue impedance seen by the EGG device is inversely proportional to lateral contact area of the vocal folds" (Childers & Krishnamurthy, 1985:133-134).

The amplitude of the signal changes because of permanently varying vocal fold contacts. It depends on:

It may happen that the impedance fluctuation caused by the vocal folds' movements is too weak to be registered (Childers & Krishnamurthy , 1985; Colton & Conture, 1990; Marasek, 1995b). It also has to be noted that EGG signals of acceptable quality are harder to obtain from woman and children than from men. This is related to the smaller mass of the vocal folds, the wider angle of the thyroid cartilage and different proportions between different types of tissue (Colton & Conture, 1990).

An integral part of the electroglottographic signal is the varying component generated by the vertical movement of the whole larynx. Therefore, the signal of rapid movements of the  vocal folds is superimposed on the signal produced by the slower movements of the other structures. Fourcin & Abberton (1977) proposed the name Gx for the waveform of larynx movement and the name Lx for the vibration component. The Gx component originates, for example, can be observed in swallowing, but it may also be caused by the vertical movement of the larynx which is related to the voice quality setting of the raised/lowered larynx. This feature is rarely considered to be informative. Fluctuations of it type are usually regarded as unusable and are removed from further analysis, despite Rothenberg and Mahshie (1988) who used Gx to calculate vocal fold abduction (this point will be dealt with in more detail further below). The DC offset changes (Gx) can be evened out because the effects of the varying larynx height are compensated by the use of additional electrodes or high pass filtering of the registered signal. The latter method may involve signal distortion, especially for low-pitched voices. The distortions may be caused by a too high cutoff frequency of the filter1 (or a too wide filter transition band). This can cause the attenuation of the Lx signal component. The non-uniform phase response function of the filter can also change the shape of the filtered waveform (the phase dependences between signal components can be altered). In principle FIR (finite impulse response) filters should be used to prevent nonlinear phase shifts in the signal components. It should be noted however, that even the unfiltered output of the EGG device is not free of distortion. Particularly the demodulation circuit whose frequency transfer function may influence the frequency response of the EGG device, especially in the low frequency range constitutes an additional source of signal shape deformation.

Many of the commercially available devices include an automatic gain control facility which is used to compensate for the variations of the signal level that are due to varying throat impedance. Such circuits respond with a small time delay and the time constant of the device may influence the EGG waveform at low frequencies (Rothenberg, 1981). Automatic gain control is however very helpful for eliminating a changed skin-electrode contact.

Figure 15. The audio and EGG signal of sustained /a/ vowel production. The basic phases of the EGG are marked additionally.

                  

Despite the problems just mentioned electroglottography has established itself as a valuable method for evaluating laryngeal behaviour. In comparison to other methods of glottography Frokjaer-Jensen observed as early as 1969 that electroglottography allows for a better representation of the closed and closing phases of vocal fold movement, especially of the vertical contact area. Photoglottography seems particularly advantageous as far as the description of the open phase is concerned. The EGG is superior to all other methods in that it is not uncomfortable to speakers as it is completely non-invasive (it exerts no influence at all on the articulation and production of sounds).