- Figure 35. The mean and standard deviation of Open Time for speaker groups
The mean of the fundamental frequency is not relevant for the distinction of the speaker groups and the perturbation coefficient of the pitch period also does not show significant differences between speaker groups. This is a rather unexpected result, as jitter is discussed in the literature (Eskénazi et al., 1990) as one of the best indicators of voice quality, especially in pathological voices. This rises the questions about the relevance of the measure used here. Perhaps, other more sophisticated methods of jitter measurement should be included. A good candidate is the jitter estimation method developed by Vieira et al. (1996b).
The ANOVA analysis of the effect of different speakers on the Speed Quotient (SQ) does not show significant results. Once again, the great variability of the results (probably caused by the imprecise determination of the opening instant) leads to non-significant relationships. It is observed however, that the mean value of the SQ for recurrent nerve paralysis patients without compensatory articulation was about three times higher than for normal voices.
The differences between the mean values of the Open Quotients of the respective speaker group are also not significant. The OQ was computed according to the first measure of the duty ratio, as defined in section 12. It appears that in the case of unusual EGG shapes (often noticed in this experiment) the method of opening instant estimation fails, resulting in a great variation of OQ results. The lower quality of the OQI measure was confirmed already in previous experiments with modal speakers (section 17.2). The Open Time (in previous experiments called Open Quotient II), measured as the relative duration of the no-contact phase, seems to be more robust with respect to pathological data. The results are shown in Fig.35.
|
The results of ANOVA with two factors (speaker group and vowel) are significant with regard to the speaker group factor (F(4,45)=7.189, p<0.0005). The post-hoc examination of the data displays the significant difference between the analyzed groups. The open phase of vocal fold vibration is substantially shorter for patients with chordectomy than it is for the other groups. As expected, the no-contact phase of the EGG lasts much longer for breathy speakers than for the other groups, except for the group of patients suffering from compensated recurrent nerve paralysis (rp-with-comp). The differences between the other groups are not significant.
The duration of the maximal contact phase (denoted as segment (b) in Fig.22 in section 12) does not show significant differences among groups, though a tendency towards a shorter (perhaps incomplete) closure is observed for chordectomy patients.
The relative durations of the rise and fall of the contact signal exhibit significant relations between speaker groups, as the time needed to achieve full contact of the folds depends on the speaker group (F(4,55)=13.295, p<0.0005). The mean for chordectomy and recurrent nerve paralysis patients is almost doubled compared to other groups (Fig.36). The longer closing gesture is obviously related to the vocal folds pathology and interestingly, the duration of the closing phase for the rp-with-comp group is almost equal to that of the control group (7% compared to 5%).
|
A similar, although weaker, relation is found for the duration of the contact fall (opening) phase (F(4,55)=13.295, p<0.0005). The corresponding values for breathy and control are higher than for the others. The chordectomy group exhibits the shortest duration. This can be caused by erroneous estimation of the closure instant or, in fact, by another shape of the waveform.
However, the mean duration of the second part of opening (defined in section 12 as the end of the opening phase) displays a reciprocal relation, since this phase lasts longer for the chordectomy group than for the other groups.
The duration of the start of the closing phase does not depend on the grouping factor.
