Acoustic correlates of vocal quality

Area of Science:

• Speech Science
• Acoustic Phonetics
• Voice Disorders

Background:

• Understanding the acoustic correlates of voice quality is crucial for diagnosing and treating vocal disorders.
• Previous research has explored various acoustic measures, but their predictive power for specific vocal qualities remains an area of investigation.

Purpose of the Study:

• To investigate the relationship between subjective voice quality ratings and objective acoustic measures in individuals with pathological voices.
• To identify the most effective acoustic parameters for predicting specific vocal qualities: overall severity, hoarseness, breathiness, roughness, and vocal fry.
• To assess whether acoustic measures can differentiate or rank normal voices.

Main Methods:

• Acoustic analysis of the vowel /i/ from normal and pathological voices using inverse filtering and linear predictive coding (LPC).
• Formal listening tests to rate pathological voices on five qualities (severity, hoarseness, breathiness, roughness, vocal fry) and normal voices on overall excellence (scale 1-7).
• Multiple linear regression analysis with prediction sums of squares (PRESS) to establish predictive equations.

Main Results:

• Useful prediction equations (order two or less) were developed relating acoustic measures to the perceived vocal qualities of pathological voices.
• Pitch Amplitude (PA) and Harmonics-to-Noise Ratio (HNR) emerged as the two most significant acoustic parameters for predicting vocal quality.
• No acoustic measure was found to be effective in ranking the overall excellence of normal voices.

Conclusions:

• Specific acoustic measures, particularly PA and HNR, can reliably predict key vocal qualities in individuals with voice disorders.
• These findings offer potential for objective assessment tools in clinical voice evaluation.
• Acoustic analysis is less effective for differentiating or ranking the subtle variations in normal voice production.