Abstract
Precise analysis of the vocal fold vibratory pattern in a stroboscopic video plays a key role in the evaluation of voice disorders. Automatic glottis segmentation is one of the preliminary steps in such analysis. In this work, it is divided into two subproblems namely, glottis localization and glottis segmentation. A two step convolutional neural network (CNN) approach is proposed for the automatic glottis segmentation. Data augmentation is carried out using two techniques : 1) Blind rotation (WB), 2) Rotation with respect to glottis orientation (WO). The dataset used in this study contains stroboscopic videos of 18 subjects with Sulcus vocalis, in which the glottis region is annotated by three speech language pathologists (SLPs). The proposed two step CNN approach achieves an average localization accuracy of 90.08% and a mean dice score of 0.65.
Highlights
IntroductionThe extensiveness rate of the Sulcus vocalis (SV) has been found to vary from 0% to 9% [3]
A dataset consisting of stroboscopic videos from 18 subjects with Sulcus vocalis (SV) are annotated by three speech language pathologists (SLPs) which is used to evaluate the proposed approach
The two step CNN WO method achieves an average localization accuracy of 90.08% and a dice score of 0.65 outperforming the baseline scheme by 24.64% and 0.26, respectively
Summary
The extensiveness rate of the SV has been found to vary from 0% to 9% [3] In this condition a groove is formed in the vocal fold, which causes a reduction in the mass of the vocal folds. This leads to an incomplete closure of the glottis during production of voice, this is termed as glottic chink. It is challenging for the SLP to quantitatively assess the glottic chink from the endoscopic video using naked eyes. This brings the need for automatic detection and segmentation of glottis from these videos. The area quantification of the segmented glottis can be used to calculate the minimal glottis opening in the video, which could assist SLPs in their assessment
Sign-in/Register to view highlights & summary 
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call