Effects of speaker voice and digit optimization on the recognition of digits in noise by normal hearing and cochlear implant listeners

Objectives: Digits-in-Noise (DIN) tests have been widely applied to assess hearing function and speech-in-noise perception, but their robustness to variations in digit stimuli between listener groups remains unknown. This study developed a DIN test for research and clinical applications and evaluated several speaker and listener factors across different listener groups, including users of cochlear implants (CI), to assess their impact on DIN test results and robustness. Methods: Audio recordings of digits (0 to 9) spoken by two British English adults (male and female) were acquired in quiet and simulated noisy backgrounds to elicit the Lombard effect. Three experiments were then performed to assess the robustness of the DIN test across several speaker and listener factors. The first experiment measured digit intelligibility in noise for single digit and triple digit stimuli in four speaker voice conditions and across native and non-native normal hearing (NH) listeners. The second experiment assessed DIN results for simulated and real CI listeners in two speaker voice conditions. The third experiment investigated the effect of digit optimization using digit level corrections for groups of NH and CI listeners in two speaker voice conditions. Results: For normal hearing listeners, speech reception thresholds (SRTs) and measurement errors were in line with previous DIN validations for all voice conditions. The different speaker voices, recording settings and language backgrounds did not significantly affect SRTs. The results of the CI group were similar between male and female speaker voices and overall comparable to previous studies, but different to the simulated CI group. Digit optimization slightly improved SRTs and test slopes for the NH groups, but not for the CI group for whom test slopes decreased with optimisation. Conclusions: The DIN test is a robust measure of auditory perception across variations in digit stimuli and listener groups. Findings were generally in line with previous studies but suggested a limited impact of digit optimization on DIN test outcomes in English, especially for CI listeners. Depending on the intended application, DIN test development may be simplified to facilitate its application to new languages and user groups for increased inclusion.