Correlation dimension of electroglottographic data from healthy and pathologic subjects

Area of Science:

• Biophysics
• Nonlinear Dynamics
• Speech Science

Background:

• Vocal fold vibration analysis is crucial for understanding voice production and pathologies.
• Correlation dimension is a nonlinear measure used to characterize dynamical systems.
• Nonstationarity, noise, and data limitations can affect dynamical system analysis.

Purpose of the Study:

• To evaluate the reliability of correlation dimension estimation from electroglottographic (EGG) signals.
• To investigate the impact of vocal fold pathologies on correlation dimension calculations.
• To determine the applicability of correlation dimension to clinical voice analysis.

Main Methods:

• Reconstruction of state space from EGG signals using the method of lags.
• Calculation of correlation dimension from EGG data of healthy and pathological subjects.
• Analysis of scaling regions and dimension values across different embedding dimensions.

Main Results:

• Correlation dimension values saturated early (between first and second embeddings) in healthy subjects and some pathological cases.
• Reliable scaling regions were identified in healthy subjects and nearly periodic pathological EGG data.
• Estimation of correlation dimension was difficult and unreliable for most pathological EGG data due to constricted scaling regions.

Conclusions:

• Correlation dimension is a subjective measure and not reliably applicable to abnormal EGG data from subjects with vocal fold pathologies.
• The presence of nonstationarity and noise in pathological EGG signals complicates accurate dimension estimation.
• Cautious interpretation and graphical presentation are recommended when using correlation dimension for EGG signal analysis.