A Fast Semiautomatic Algorithm for Centerline-Based Vocal Tract Segmentation.

Vocal tract morphology is an important factor in voice production. Its analysis has potential implications for educational matters as well as medical issues like voice therapy. The knowledge of the complex adjustments in the spatial geometry of the vocal tract during phonation is still limited. For a major part, this is due to difficulties in acquiring geometry data of the vocal tract in the process of voice production. In this study, a centerline-based segmentation method using active contours was introduced to extract the geometry data of the vocal tract obtained with MRI during sustained vowel phonation. The applied semiautomatic algorithm was found to be time- and interaction-efficient and allowed performing various three-dimensional measurements on the resulting model. The method is suitable for an improved detailed analysis of the vocal tract morphology during speech or singing which might give some insights into the underlying mechanical processes.

Lower Vocal Tract Morphologic Adjustments Are Relevant for Voice Timbre in Singing.

The vocal tract shape is crucial to voice production. Its lower part seems particularly relevant for voice timbre. This study analyzes the detailed morphology of parts of the epilaryngeal tube and the hypopharynx for the sustained German vowels /a/, /e/, /i/, /o/, and /u/ by thirteen male singer subjects who were at the beginning of their academic singing studies. Analysis was based on two different phonatory conditions: a natural, speech-like phonation and a singing phonation, like in classical singing. 3D models of the vocal tract were derived from magnetic resonance imaging and compared with long-term average spectrum analysis of audio recordings from the same subjects. Comparison of singing to the speech-like phonation, which served as reference, showed significant adjustments of the lower vocal tract: an average lowering of the larynx by 8 mm and an increase of the hypopharyngeal cross-sectional area (+ 21:9%) and volume (+ 16:8%). Changes in the analyzed epilaryngeal portion of the vocal tract were not significant. Consequently, lower larynx-to-hypopharynx area and volume ratios were found in singing compared to the speech-like phonation. All evaluated measures of the lower vocal tract varied significantly with vowel quality. Acoustically, an increase of high frequency energy in singing correlated with a wider hypopharyngeal area. The findings offer an explanation how classical male singers might succeed in producing a voice timbre with increased high frequency energy, creating a singer`s formant cluster.

Tonotopic morphometry of the lamina reticularis of the guinea pig cochlea with associated microstructures and related mechanical implications.

Morphometry of the lamina reticularis of the guinea pig cochlea was performed using scanning electron microscopy. Seventy-four geometrical parameters of the lamina reticularis, the bundles of stereocilia, and individual stereocilia, in all rows of hair cells and within the individual hair cells, were measured at ten equally spaced locations along the longitudinal direction of the cochlea. Variations of the parameters versus the longitudinal coordinate were statistically analyzed and fitted with polynomials (constant, linear, or quadratic). Our data show that a unique set of geometrical parameters of inner and outer hair cells is typical for every frequency-dependent position at the lamina reticularis. Morphology of the outer hair cell structures varies more than respective parameters of the inner hair cells. Mechanical modeling using the obtained geometrical parameters provides a novel glance at the mechanical characteristics with respect to the cochlear tonotopy.

A segmentation method to obtain a complete geometry model of the hearing organ.

We present a method for obtaining a complete geometry model of the fluid chambers of cochlea (scalae) from tomography images. An accurate segmentation of cochlea is problematic due to the low contrast of the inner membranes of scalae. Our method of 3D segmentation is based on dynamic resampling of an original image stack to achieve a perpendicular cross-section of the scalae on all sections. Subsequently, perpendicular cross-section is being segmented using 2D active contours. The center of mass of the contour is extracted and used to predict further course of scalae centerline by Kalman filter. Cross-section contours are subsequently assembled to the total geometry model. This method has been applied to CT images, but we expect that it could be used for segmentation of strongly curved low-contrast tubular objects recorded with other tomography techniques.

The creation of geometric three-dimensional models of the inner ear based on micro computer tomography data.

The modeling of the mechanical process of hearing requires an accurate geometrical model of the inner ear (cochlea). The purpose of this study was the creation of a 3-D model of the fluid chambers of Guinea pig cochlea, which could serve as a basis for further mechanical modeling. Micro computer tomography used in this study is a noninvasive method to visualize bony structures. The visualization of the membranous labyrinth was achieved by additional staining of the specimen with OsO(4). The resulting stack of images has been transformed into a cylindrical coordinate system. To suppress noise on tomography images, a nonlinear smoothing method, anisotropic diffusion, were applied. A new approach has been proposed to estimate algorithm parameters automatically. Then, a segmentation using active contours (snakes) was performed. In this study, a new energy linking the contours on adjacent slices has been added to the standard approach. This compensates the inconsistencies between adjacent contours. The images segmented in this way were used as a basis for a 3-D reconstruction of the hearing organ.