Back to main page

A review of methods used for the analysis of the electroglottographic signal

Methods relying on the derivative of the electroglottographic signal (DEGG signal)

1. Detection of the peaks by applying thresholds to the DEGG signal

These methods consist in detecting the maxima located above a given positive threshold and the minima located below a given negative threshold, as illustrated in case (d) in the following Figure. This threshold can be chosen as a percentage of the maximum (respectively minimum) amplitude of the windowed signal. Several options are open for peak detection. For instance, one may look for the local minimum in-between consecutive glottal closing instants, or opening peaks may be detected over the whole of the windowed signal and then matched with closing peaks.

2. Correlation-based method

The DECOM (DEgg Correlation-based Open quotient Measurement; Henrich et al., 2004) method is applied to a four-period windowed DEGG signal which is separated into two parts: its positive part, which shows strong peaks related to glottal closing instants, and
its negative part, which shows weaker peaks related to glottal opening instants. The fundamental period duration is derived from the autocorrelation function calculated on the positive part of the DEGG signal. The open time is derived from the intercorrelation function calculated between the positive part and the negative part.

Threshold methods applied directly to the EGG signal

In the literature, the main electroglottographic-based methods for detecting the glottal closing and opening instants use the EGG signal. As illustrated by cases (a) and (b) in the following Figure, a threshold method is generally applied (Rothenberg and Mahshie, 1988): a level line is chosen either as a percentage of the amplitude between the minimum and maximum of the signal over a glottal period (50 % for a normal to pressed voice and 35 % for a relaxed voice).


Illustration of 4 methods for measuring the open quotient (or its equivalent, the closed quotient) on
an EGG signal and its derivative: a) on the EGG signal using a threshold detection (35 % of the difference between the minimum and maximum value of the signal over a glottal period); b) same method as a) with a threshold of 50 % ; c) in combining an EGG-based threshold method (3/7) with a detection of glottal closing instants on the DEGG signal; d) on the DEGG signal by the detection of opening and closing peaks. From Henrich et al., 2004.

The crossing points between the level line and the EGG signal are taken as approximations of the instants of glottal closing and opening and these are used for the open quotient measurement. These methods are very convenient for medical purposes, as they are robust and can be applied even on noisy or weak signals. Nevertheless, the results of such methods are by their nature imprecise and can be inaccurate, as compared to what would be measured on a glottal area signal or on a glottal flow signal (Rothenberg and Mahshie, 1988).

A method combining detection of closing instants on the DEGG signal and opening instants on the EGG signal

An improvement over EGG threshold methods has been proposed by Howard (1995); Howard et al. (1990): using the DEGG closing peak for detecting the glottal closing instant, and an EGG-based threshold method for detecting the glottal opening instant. This is illustrated by case (c) in the Figure above.

[Note: the references are on the main page.]

Back to top