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

Area of Science:

• Laryngology
• Speech Science
• Biomedical Engineering

Background:

• High-speed videoendoscopy (HSV) captures laryngeal dynamics during speech.
• Temporal segmentation of HSV data is crucial for voice analysis.
• Current methods may lack automation for connected speech.

Purpose of the Study:

• To develop and validate a gradient-based method for temporal segmentation of laryngeal HSV data.
• To enable automated mapping of vocal fold vibrations during connected speech.
• To assess the clinical potential for voice assessment protocols.

Main Methods:

• Utilized a custom HSV system with a fiberoptic nasolaryngoscope.
• Developed a gradient-based algorithm to create a motion window for tracking vocal folds.
• Estimated glottal area waveform and computed fundamental frequency (f0) using image processing and autocorrelation.
• Performed temporal segmentation based on f0 contour and epiglottic obstruction detection.

Main Results:

• Cross-validated automatic and visual temporal segmentation methods.
• Demonstrated agreement between automatic f0-contour patterns and visual inspection of vibratory frequency changes.
• Successfully mapped vocalization onsets, offsets, and epiglottic obstructions.

Conclusions:

• Confirmed the feasibility of automatic temporal segmentation for laryngeal HSV during connected speech.
• The automated analysis provides valuable insights into vocal fold dynamics.
• This method holds significant potential for advancing instrumental voice assessment.