Collection and Analysis of Repeated Speech Samples: Methodological Framework and Example Protocol

Area of Science:

• Speech and Language Pathology
• Biomedical Engineering
• Computational Linguistics

Background:

• Speech and language biomarkers offer potential for objective health assessments but face challenges due to high within-individual variability.
• Inconsistent reporting of speech characteristics in research hinders methodological progress, replication, and biomarker identification.
• Existing methodologies often fail to adequately account for confounding variables, impacting the reliability of speech-based health assessments.

Purpose of the Study:

• To present a transparent speech collection and feature extraction protocol and design checklist to facilitate replicable speech research.
• To demonstrate the feasibility of an in-laboratory protocol designed to minimize confounding factors in healthy speech recordings.
• To establish a standardized framework for collecting and analyzing speech data for health assessment applications.

Main Methods:

• Developed a speech collection and feature extraction protocol based on a literature review for controlled investigation of within-individual speech variability.
• Protocol includes elicitation of read speech, held vowels, and picture descriptions, extracting 14 health-relevant features.
• Speech data collected using a condenser microphone, 3 smartphones, and a headset for device-specific sensitivity analysis.

Main Results:

• Collected healthy speech data from 28 individuals (day cohort) and 25 individuals (week cohort) with repeated measures.
• Collected participant characteristics (sex, age, native language) and contextual data (voice use, intake, emotional state, recording times).
• Analysis of within-individual variability and device-type sensitivity is ongoing, with results anticipated in 2025.

Conclusions:

• Variability in speech data collection, processing, and reporting impedes clinical translation of speech biomarkers.
• A speech curation protocol design checklist is developed to address inconsistencies in research.
• Standardized reporting and justified protocols are crucial for enabling speech research replication and clinical application.