Spatial Segmentation for Laryngeal High-Speed Videoendoscopy in Connected Speech.

Area of Science:

• Biomechanics of speech production
• Computational imaging
• Laryngology

Background:

• High-speed videoendoscopy (HSV) provides detailed visualization of vocal fold dynamics.
• Accurate segmentation of vocal fold edges is crucial for quantitative voice analysis.
• Existing methods may struggle with the complexities of connected speech.

Purpose of the Study:

• To develop an automated computational framework for spatial segmentation of vocal fold edges in HSV data during connected speech.
• To enable spatio-temporal analytic representation of vocal folds for vibratory characteristic measurement.
• To facilitate HSV-based measurement of the glottal area waveform in running speech.

Main Methods:

• Developed an algorithm based on active contour modeling for HSV data analysis.
• Applied the algorithm to HSV kymograms to detect vibrating vocal fold edges.
• Used an energy optimization procedure with deformation rules for edge detection.
• Registered detected edges back to HSV frames to calculate the glottal area waveform.

Main Results:

• The algorithm successfully captured vocal fold edges in HSV kymograms.
• Automated measurement of the glottal area waveform from HSV frames during connected speech was achieved.
• Demonstrated the feasibility of analyzing vocal fold dynamics in running speech.

Conclusions:

• The proposed algorithm provides an automated method for vocal fold spatial segmentation in HSV data during connected speech.
• This represents an initial step towards developing HSV-based measures for studying vocal fold vibratory characteristics.
• The framework aids in understanding voice production mechanisms in both normal and disordered voices during connected speech.