JMIR Publications

ABSTRACT

Background:

Speech and language biomarkers have the potential to be regular, objective assessments of symptom severity in several neurological and mental health conditions, both in-clinic and remotely using mobile devices. However, the complex nature of speech and often subtle changes associated with health mean that findings are highly dependent on methodological and cohort choices. These are often not reported adequately in studies investigating speech-based health assessment, which (i) hinders the progress of methodological speech research, (ii) prevents replication, and (iii) makes the definitive identification of robust biomarkers problematic.

Objective:

1. To facilitate replicable speech research by presenting an adaptable speech collection and analytical method and design checklist for other researchers to adapt for their own experiments. 2. To develop and apply an exemplar protocol that reduces and controls for confounding factors in repeated recordings of healthy speech, including device choice, speech elicitation task and non-pathological variability.

Methods:

We developed a collection protocol based on a thematic literature review. Our protocol comprises the elicitation of (1) read speech, (2) held vowels, and (3) a picture description. With a focus towards remote applications, we collected speech with different devices: a freestanding condenser microphone, three smartphones and a headset. We developed and report in detail a pipeline to extract a set of 14 exemplar speech features, also chosen via a thematic literature review, that cover timing, prosodic, quality, articulatory, and spectral characteristics of speech.

Results:

We collected healthy speech from 28 individuals three times in one day, repeated at the same times 8-11 weeks later, and from 25 individuals on three days in one week at fixed times. Participant characteristics collected included sex, age, native language status, and the participant’s voice use habits. Before each recording, we collected information on recent voice use, food and drink intake, and emotional state. Recording times were also documented. Values of the extracted features are presented providing a resource of normative values.

Conclusions:

Speech data collection, processing, analysis and reporting towards clinical research and practice varies widely, motivating this report and speech collection protocol design checklist. Greater harmonization of study protocols and consistent reporting are urgently required to translate speech processing into clinical research and practice.