Measuring Talker Age Estimates Through Crowdsourced Listeners' Ratings: A Pilot Study for Voice Research
Raquel M A Tripp1, Eric J Hunter2, Aaron M Johnson3
1Department of Communicative Sciences and Disorders, New York University, NY.
Purpose:
Most auditory-perceptual voice research utilizes the judgments of trained listeners rather than everyday listeners with no previous training in speech pathology. Online crowdsourcing of behavioral data from untrained participants is rapidly increasing in popularity but has yet to be a common procedure for auditory-perceptual studies of the voice. The objective of this pilot study was to assess the functionality of this model for judgments of voice by using an online experiment platform to replicate a lab-based, voice-specific age estimation study.
Method:
Fifty crowdsourced untrained listeners estimated the age of a single talker based on audio samples taken from 20 speeches over a 48-year span. The primary outcome was overall age estimation accuracy.
Results:
The crowdsourced age estimations closely matched those of a previous highly controlled in-person laboratory study using the same auditory samples. Listeners generally overestimated the talker's age when the talker was younger and underestimated his age when he was older. The age at which the estimated age equaled the talker's chronological age was 54 years.
Conclusions:
Online crowdsourcing may be a feasible modality for auditory-perceptual voice ratings with the potential to add low-cost, high-number options to validate and enhance clinical and laboratory-based studies by (a) including a wider diversity of participants and (b) providing the means for rapidly recruiting more participants. Further research investigating crowdsourced ratings of the complex parameters of voice quality using more listeners is needed to continue supporting this methodology as a tool for perceptual voice research.
