Validation of the French-language version of the OTOSPEECH automated scoring software package for speech audiometry.

OBJECTIVES To validate a novel speech audiometry method using customized self-voice recorded word lists with automated scoring. PATIENTS AND METHODS The self-voice effect was investigated by comparing results with prerecorded or self-recorded CVC (consonant-vowel-consonant) word lists. Then customized lists of 3-phoneme words were drawn up using the OTOSPEECH software package, and their scores were compared to those for reference lists. Finally, the customized list scores were compared on automated (Dynamic Time Warping [DTW]) versus manual scoring. RESULTS Self-voice did not change scores for perception of CVC words at 10, 20 and 30 dB (ANOVA>0.05). Scores obtained with pre-recorded and self-recorded lists correlated (n=10, R(2)=0.76, P<0.01). Customized list scores correlated strongly with the reference cochlear lists of Lafon in normal-hearing (n=77, R(2)=0.83, P<0.001) and hearing-impaired populations (n=13, R(2)=0.89, P<0.001). Results on the automated and manual scoring methods correlated in both populations (n=77, R(2)=0.71, P<0.01; and n=13, R(2)=0.76, P<0.01, respectively), with DTW scores ranging from 24.17 to 53.24. CONCLUSIONS Automated scoring of customized self-voice recorded lists for speech audiometry displayed results similar to conventional audiometric techniques.