Biomedical Image Analysis in High-speed Laryngeal Imaging of Voice Production

High-speed laryngeal imaging (HSLI) is emerging as a new clinical tool for the assessment of voice disorders. The analysis of high temporal resolution images of the vibrating vocal folds may provide new information on the mechanism of voice production and how altered vibratory properties relate to voice pathologies. An essential procedure in the HSLI data processing is tracing the vocal fold edges, an operation referred to as glottal edge detection. In this paper, we present an adaptive image segmentation method which facilitates fast and effective glottal edge detection operating on a frame-by-frame basis. The glottal area waveform traced from this process is then analyzed using an analytic signal approach we previously developed. In this approach, the analytic phase trace plot, or refereed to as Nyquist plot, is used to show patterns of the vocal fold vibration that correlate with voice quality and health condition. Several examples of clinical voice recordings are used to evaluate and test the effectiveness of our approach.