In addition to pure-tone audiometry tests and electrophysiological tests, a comprehensive hearing evaluation includes assessing a subject’s ability to understand speech in quiet and in noise. In fact, speech audiometry tests are commonly used in clinical practice; however, they are time-consuming as they require manual scoring by a hearing professional. To address this issue, we developed an automated speech recognition (ASR) system for scoring subject responses at the phonetic level. The ASR was built using a deep neural network and trained with pre-recorded French speech materials: Lafon’s cochlear lists and Dodelé logatoms. Next, we tested the performance and reliability of the ASR in clinical settings with both normal-hearing and hearing-impaired listeners. Our findings indicate that the ASR’s performance is statistically similar to manual scoring by expert hearing professionals, both in quiet and in noisy conditions. Moreover, the test–retest reliability of the automated scoring closely matches that of manual scoring. Together, our results validate the use of this deep neural network in both clinical and research contexts for conducting speech audiometry tests in quiet and in noise.
Understanding speech in noise is particularly difficult for individuals occupationally exposed to noise due to a mix of noise-induced auditory lesions and the energetic masking of speech signals. For years, the monitoring of conventional audiometric thresholds has been the usual method to check and preserve auditory function. Recently, suprathreshold deficits, notably, difficulties in understanding speech in noise, has pointed out the need for new monitoring tools. The present study aims to identify the most important variables that predict speech in noise understanding in order to suggest a new method of hearing status monitoring. Physiological (distortion products of otoacoustic emissions, electrocochleography) and behavioral (amplitude and frequency modulation detection thresholds, conventional and extended high-frequency audiometric thresholds) variables were collected in a population of individuals presenting a relatively homogeneous occupational noise exposure. Those variables were used as predictors in a statistical model (random forest) to predict the scores of three different speech-in-noise tests and a self-report of speech-in-noise ability. The extended high-frequency threshold appears to be the best predictor and therefore an interesting candidate for a new way of monitoring noise-exposed professionals.
Understanding speech in noise is particularly difficult for individuals occupationally exposed to noise due to a to a mix between noise-induced auditory lesions and energetic masking of speech signal. For years, the monitoring of conventional audiometric thresholds has been the usual way to check and preserve the auditory function. But the highlighting of supra-threshold deficits, notably speech in noise understanding difficulties, has pointed out the need for new monitoring tools. The present study aims to identify the most important variables that predict speech in noise under-standing in order to suggest a new way of regular monitoring. Physiological (distortion products of otoacoustic emissions, electrocochleography) and behavioral (amplitude and frequency modula-tion detection thresholds, conventional and extended high frequency audiometric thresholds) variables were collected in a population of individuals presenting a relative homogeneous occu-pational noise exposure. Those variables were used as predictors in a statistical model (random forest) to predict scores of three different speech in noise tests and a self-report of speech in noise ability. The extended high frequency threshold appears to be the best predictor and therefore an interesting candidate for a new way of monitoring noise exposed professionals.