11. The EGG waveform and voice qualities

Figure 21. Examples of EGG recordings of different voice qualities

The dependence of the EGG waveform on voice quality is evident in Fig.21. Several periods of Lx recordings of /a:/ spoken by the same speaker are compared using the same amplitude and time resolution settings. The visual inspection of the waveforms leads to the following conclusions:

• 
• Modal voice is characterized by a steep increase of contact in the closing phase. The closing phase is short and peak-to-peak amplitude is high. The maximum contact phase (closed phase) has a parabolic shape and the transition to the rise⁷ of the impedance is smooth. The fall of the signal lasts longer than the rise (the skewness is substantially smaller than 100%) and in the increase of the impedance a knee can be observed. The increase in impedance ends at the minimum located at the beginning of the no-contact phase, from then on the impedance slowly decreases until the sharp decrease at the beginning of the next signal period. The signal duty ratio is about 50% (the Open Quotient), although the location of the opening instant is not obvious. The waveform directly corresponds to the vocal folds vibration model and its EGG projection described in section 26 (Fig. 45). Moreover, the phases of the signal are almost linear (exept for the maximum contact phase).
• 
• Whispery voice periods are slightly longer than those of modal voice. The decrease in impedance is very fast. The signal peak is located at the beginning of the maximum contact phase which is much shorter than for modal voice. The increase in impedance is also faster and the knee in the fall is present, although in this particular example in would be preferable if the opening instant were located at the signal level of the closure instant (i.e. where the falling edge of the signal changes from concave to convex, not where the time derivative reaches the minimum). The skewness of the contact pulse is still much smaller than 100% but still larger than for modal voice. The minimum of the signal amplitude is reached at about the beginning of the no-contact phase. From this point on the waveform rises linearly up to the discontinuity point at the begining of the sharp rise which marks the next period of the signal. The duty ratio is much lower than for modal voice, indicating a much higher Open Quotient. The high amplitude of the signal suggests that there is complete adduction of the mucous part of the vocal folds. Once again, the phases of the EGG waveform are easy to identify in the EGG model. It is to be expected that the durations of the single phases differ from those of modal voice.
• 
• The creaky voice recordings of the EGG contrast in many aspects with the other voice qualities. The shape of the waveform could be described as a "triangle with rounded corners" and the period duration is ca. twice as long as for modal voice. The instant of glottal closure can be identified clearly at the sharp rise of the waveform. The maximum contact is short compared to the period duration. In the opening phase the signal gradually decrease and no knee can be identified. Due to this fact, the signal duty cycle is not easy to define and it is to be expected that the measure will be biased and imprecise. With respect to the skewness of the pulse, one can observe a very strong tendency to the left, presumably higher than for the other voices. The no-contact phase has a rather parabolic form with the signal minimum located at the centre of the phase. Using the EGG model (see section 9) the description of a waveform is more complicated due to the undetectable opening instant. The start and end of the opening phases (see section 12) are approximated with two straight lines with almost the same slope. Also, the initial part of the closing phase with a rather gradual increase in the signal is difficult to localize. The peak-to-peak amplitude is still high, which signalizes a good current flow during the contact phase, which in turn suggests complete contact between the vocal folds. Noise is also present in the signal.
• 
• Breathy voice is easy to distinguish from the other voices presented in Fig.21. First of all, the peak amplitude is lower, which may depend on the poor contact between the vocal folds or their incomplete closure. Also, the duty ratio of the signal is obviously greater than in other voice qualities. Considering the ratio of rise time and fall time, higher values are observed. The skewness of the pulse is smaller. The maximum contact phase is extremely short with the maximum located approximately in the middle of the pulse. The contour of the waveform can be described as consisting of short triangular pulses with long, straight, almost zero-valued, flat pauses (plateaus) between them. The instants of glottal closure and opening can be identified as intersections with the baseline. In the no-contact phase, a slow fluctuation of the signal can be observed. The description of the waveform using the proposed model is still possible, although some phases (for example the start of closing or the end of opening) are obviously reduced or even absent. The pitch is low but higher than for creaky voice.
• 
• Tense voice (in Fig. 21) is produced with a strongly increased muscular tension in the larynx. The EGG waveform typical of tense voice differs in many aspects from the waveforms of other voice qualities. First of all, the shape is rounded, much more sinusoidal and smoothed when compared to other voices. The peak-to-peak amplitude is lower, which can be caused by incomplete glottal closure. The steepness of the increasing signal slope is significantly reduced and the parabolic part starts earlier and lasts longer during the maximum contact phase. The increase in the impedance is gradual with a slight indication of a knee, but it seems that the use of this time instant as an estimate of the opening instant is very unreliable. The maximum contact phase is relatively long. The duty cycle of the signal is comparable to that of modal voice. The skewness is rather high, but the rise time relatively long. The minimum of the signal is reached rather late in the no-contact phase. The fundamental frequency of the signal is comparable to that of modal voice, whereas the signal-to-noise ratio seems to be lower. The description of tense voice in terms of the proposed model of straight lines (section 12) is biased due to the rounded waveform. However, the measure of the distance between the modelled and the original waveform should be notably higher than for other voices. The start of the closing phase is not detectable in the waveform.
• 
• The EGG shape of falsetto also differs from the other voice qualities. The pitch periods are almost three times shorter than for modal voice, the minimum and maximum peaks are sharp and the waveform can be accurately approximated with straight lines. The rise of the signal is extremely fast. The maximum contact is of very short duration. The fall is rapid, no knee is observed, although in the middle part of the fall the time derivative reaches its minimum. The position of the opening instant is uncertain. The skewness of the pulse is limited and so the pulses are much more symmetrical than in the other examples. The EGG signal reaches the minimum relatively early in the no-contact phase and quasi-linear and fast rise can be observed up to a sharp acceleration at the closure instant. The approximated duty cycle is rather high when compared to whispery voice.

  It is apparent from the depicted waveforms that the EGG differs for diverse voice qualities. Below, an objective and automated way of describing those differences will be presented. The proposed description of the EGG waveform will be used to distinguish voice qualities that are used on the linguistic and paralinguistic layers of communication.