Abstract
Until now, it has been impossible to discriminate a pathology of the vocal folds and, in many instances, even to distinguish normal from pathological voices with an electroglottographic signal (EGG). To introduce a method for analyzing electroglottographic signals and for extracting features able to characterize phonation quantitatively. The EGG signal recorded during a continuous vocal phonation is processed in order to obtain the first derivative, which is related to the velocity of movements and contact of the vocal folds. The average fundamental frequency is computed and its corresponding period is taken as the typical duration of the EGG cycle. After each glottal cycle has been identified, the EGG signal and its derivative are locally normalized in time. For each glottal cycle, the amplitude and related velocity signals are plotted in an X-Y graph thus forming a multi-layer display where each EGG cycle appears as a circular trace. This X-Y representation can be viewed as a polar graph: by increasing the angle from 0 to 360° with incremental steps corresponding to the time normalization re-sampling of the EGG cycle, mean value and variance are computed. The results are the curve of the amplitude-velocity mean cycle and the related variance curve. The shape of the mean loop is strictly associated with the relationships between amplitude-velocity changes and phonation phases. The surrounding area represents the variability of local vocal phenomena around the above mean curve. The phonation process can be characterized in more detail by computing couples of indices (mean and variance) as obtained by dividing the polar graph in 4 quadrants, roughly associated with the different phases of the glottal cycle. In our study we carried out the EGG analysis of 21 cases of normal voice and 21 cases of pathological voice, considering the variability based on the combined amplitude-velocity analysis. In normal subjects, the global variability indices (VI) (expression of Amplitude and Velocity variation) and the four VI of different physiological phases of glottal wave (VI1, VI2, VI3 and VI4), were definitely lower than in pathological subjects. Such difference was statistically significant (p < 0.03). The above method for analyzing the EGG signal proved to be efficient to discriminate normal subjects from pathological ones. Additional trials with more subjects are needed to confirm this preliminary data and to evaluate possible differences between different pathologies.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have