Improving reliability and accuracy of vibration parameters of vocal folds based on high-speed video and electroglottography

Quantified vibration parameters of vocal folds, including parameters directly extracted from high-speed video (HSV) and electroglottography (EGG), and inverse parameters based on models, can accurately describe the mechanism of phonation and also classify the abnormal in clinics. In order to improve the reliability and accuracy of these parameters, this paper provides a method based on an integrated recording system. This system includes two parts: HSV and EGG, which can record vibration information of vocal folds simultaneously. An image processing approach that bases on Zernike moments operator and an improved level set algorithm is proposed to detect glottal edges at subpixel-level aiming at image series recorded by HSV. An approach is also introduced for EGG data to extract three kinds of characteristic points for special vibration instants. Finally, inverse parameters of vocal folds can be optimized by a genetic algorithm based on the experimental vibration behaviors synthesized with these parameters and the simulations of a two-mass model. The results of a normal phonation experiment indicate that the parameters extracted by this method are more accurate and reliable than those extracted by general methods, which were only on the basis of HSV data and with pixel-level processing approaches in former studies.