Automated Analysis of Relative Fundamental Frequency in Continuous Speech: Development and Comparison of Three...

Area of Science:

• Speech Science
• Acoustic Phonetics
• Voice Research

Background:

• Relative fundamental frequency (RFF) quantifies laryngeal tension and vocal effort during speech.
• Current RFF analysis from continuous speech necessitates manual processing, limiting large-scale ecological studies.
• Automated methods are needed to analyze RFF in continuous speech efficiently.

Purpose of the Study:

• To develop and evaluate three fully automated pipelines for RFF analysis from continuous speech.
• To address the limitations of manual processing in current RFF derivation methods.
• To enable time-efficient RFF analysis for large datasets.

Main Methods:

• Compared two modified automated relative fundamental frequency (aRFF)-AP pipelines with a novel pipeline (aRFF-B) replicating manual analysis.
• Tested pipelines on vowel-consonant-vowel (VCV) utterances from 82 female participants with and without vocal fatigue.
• Validated automated RFF measurements against manual analysis for reliability and accuracy.

Main Results:

• All three automated pipelines demonstrated good reliability (r ≥ 0.84) and validity compared to manual analysis.
• Minimal manual correction (<4%) was required, primarily for fricative identification.
• The novel aRFF-B pipeline showed the lowest sample rejection rate (10%-25%) while maintaining high reliability and enabling parallel computing.

Conclusions:

• Automated pipelines, particularly aRFF-B, allow for time-efficient RFF analysis of extensive continuous speech data.
• These advancements remove the need for manual intervention, facilitating large-scale voice studies.
• The developed pipelines can expand the application of RFF in voice research and clinical practice.