Authors: David Meng¹
¹National Acoustic Laboratories
Background: Hindered by language barriers, speech testing is not routinely included in standard clinical assessments. Although it is a valuable tool for identifying hearing challenges and informing appropriate hearing aid amplification, its limited use presents a significant challenge to delivering effective and equitable hearing care—particularly for culturally and linguistically diverse (CALD) populations.
Objective: This research presents the development and validation of a novel, language-agnostic speech test powered by artificial intelligence (AI). It employs an easy-to-administer audiovisual paradigm and test materials based on carefully selected Vowel-Consonant-Vowel (VCV) syllables.
Methods and Results: An initial validation study involving 12 normal-hearing participants and 25 hearing aid users showed that, compared to a standard Bamford-Kowal-Bench (BKB)-style sentence test, the new test demonstrated comparable test-retest reliability and sensitivity to hearing aid benefit, while requiring only approximately half the testing time under similar task demands. Subsequent refinements included optimizing the stimulus set to 57 VCV syllables and adopting an intuitive 3×2 visual response matrix. Crucially, a novel machine learning algorithm was introduced to efficiently estimate 50% speech recognition thresholds (SRT50) while capturing detailed vowel and consonant confusion patterns. Further validation with 60 normal-hearing participants from diverse linguistic backgrounds (20 each in English, Mandarin, and Greek) confirmed the test’s reliability, true language independence, and even faster administration time.
Conclusion: These findings support the potential clinical application of this easy Language-Independent Test (eLIT) for delivering efficient, equitable, and personalized hearing care across diverse populations. Moreover, the intuitive audiovisual test paradigm opens up promising opportunities for future applications in telehealth and remote hearing assessments.
