Challenges and Limits in Explaining and Acoustic Modeling of Voice Characteristics

Area of Science:

• Speech Science
• Acoustic Phonetics
• Speech Therapy

Background:

• Subjective voice assessments lack consistency and objective measures.
• Accurate voice analysis is critical for speech therapy outcomes.
• Understanding the link between acoustic features and perceived voice quality is challenging.

Purpose of the Study:

• To investigate challenges in novice speech therapists' voice assessments.
• To identify acoustic features predicting perceived voice characteristics.
• To explore the influence of speaker gender on voice perception and acoustic correlates.

Main Methods:

• Study 1: 22 trainees assessed 20 voices pre/post expert explanation.
• Study 2: 295 voice samples analyzed for acoustic features predicting 20 expert-rated characteristics.
• Statistical modeling used to link acoustic features with perceived voice traits.

Main Results:

• Expert explanation did not improve overall trainee rating performance.
• Unique acoustic features significantly predicted individual voice characteristics, with modest model fits.
• Speaker gender significantly interacted with acoustic features in predicting voice characteristics.

Conclusions:

• Novice voice practitioners face specific challenges in voice characterization.
• Objective acoustic features can predict perceived voice characteristics, but gender must be considered.
• Acoustic correlates of voice characteristics rated consistently by humans show higher predictive accuracy.