Automatic tracing of vocal-fold motion from high-speed digital images

Abstract

Defining characteristics of the phonatory vocal fold vibration is essential for studies that aim to understand the mechanism of voice production and for clinical diagnosis of voice disorders. The application of high-speed digital imaging techniques to these studies makes it possible to capture sequences of images of the vibrating vocal folds at a frequency that can resolve the actual vocal fold vibrations of a patient. The objective of this study is to introduce a new approach for automatic tracing of vocal fold motion from image sequences acquired from high-speed digital imaging of the larynx. The approach involves three process steps. 1) Global thresholding--the threshold value is selected on the basis of the histogram of the image, which is assumed to follow Rayleigh distribution; 2) applying a morphology operator to remove the isolated object regions; 3) using region-growing to delineate the object, or the vocal fold opening region, and to obtain the area of the glottis; the segmented object obtained after global threshold and the morphological operation is used as a seed region for the final region-growing operation. The performance, effectiveness and validation of our approach is demonstrated using representative, high-speed imaging recordings of subjects having normal and pathological voices.