Comparative assessment of electroglottographic and acoustic measures of jitter in pathological voices

Jitter, or the amount of cycle-to-cycle variation in the fundamental frequency, is a characteristic of the vocal folds' vibration that may affect electroglottographic (EGG) and acoustic signals in similar ways, because the translaryngeal conductance and the oral pressure are modulated by the same physiological mechanism. Despite the apparent simplicity in jitter computation and the relative facility in recording and analyzing EGG signals, only a few studies comparing EGG and acoustic jitter have been reported. This can be attributed to the remarkable sensitivity of measures of acoustic jitter to such features as the type of sound being analyzed, the equipment used for data acquisition, or the algorithms used to identify glottal cycle boundaries. To assess the agreement achievable by measures from these signals in pathological voices, acoustic and EGG waveforms of sustained vowels (/i/, /a/, /u/) produced by 15 dysphonic patients were analyzed by jitter extraction methods on the basis of peaks, zero crossings, and a waveform matching technique. The agreement, expressed as normalized absolute differences between acoustic- and EGG-derived jitter, was poorer for /i/ and /u/ than for /a/ vowels. For /a/ vowels, a method of acoustic jitter estimation is also proposed that combines peaks and zero crossings and resulted in increased consistency with the zero crossing-based EGG measures (mean normalized absolute difference: 10.95%, SD: 6.44%; range: 23.81%). The proposed method, which has a built-in-test intended to reject unreliable estimates, may provide more confidence in acoustic measures in dysphonic speakers and lead to a better understanding of the relationships between acoustic and EGG signals.