Analysis of vocal disorders in a feature space

Area of Science:

• Medical acoustics
• Signal processing
• Chaos theory

Background:

• Vocal disorder classification is crucial for clinical applications.
• Existing methods may lack precision in differentiating voice pathologies.
• Feature space analysis offers potential for improved diagnostic accuracy.

Purpose of the Study:

• To develop and validate a method for classifying vocal disorders using geometric signal separation.
• To introduce a quantitative 'healthy index' for assessing voice disorder severity.
• To demonstrate the effectiveness of the proposed method in distinguishing normal from disordered phonation.

Main Methods:

• Geometric signal separation in a feature space.
• Analysis of conventional and chaos theory metrics (entropy, correlation dimension, Lyapunov exponent, autocorrelation, spectral factor) for feature vector creation.
• Calculation of a 'healthy index' based on cluster distances in the feature space.

Main Results:

• Successful application of geometric signal separation for vocal disorder classification.
• The proposed 'healthy index' effectively quantifies the degree of vocal disorder.
• High accuracy in distinguishing between normal and disordered voice samples.

Conclusions:

• Geometric signal separation provides a robust framework for vocal disorder classification.
• The 'healthy index' offers a valuable tool for clinical assessment of voice quality.
• This approach holds promise for enhancing the diagnosis and management of vocal pathologies.