Prediction of Software-Based Sentence-in-Noise Recognition Thresholds of Hearing-Impaired Listeners

Abstract

Purpose: Although hearing-impaired (HI) listeners often have difficulty understanding in noise as their primary complaints, the speech-in-noise intelligibility test is not conducted as a standard audiologic test battery. This study investigated whether the speech audiometry in quiet accurately reflects the sentence-in-noise intelligibility of HI listeners. Methods: Sixty-two HI listeners participated. All the HI listeners had symmetrical high-frequency hearing loss and bilaterally worn hearing aids. Twenty-five normal-hearing (NH) listeners also participated as a control group. The unaided word and sentence recognition scores (WRS and SRS) were obtained in quiet at individually determined most comfortable loudness level. With bilateral hearing aids, the aided WRS and SRS were evaluated at a normal conversational level. The software-based Korean Matrix sentence in noise test was administered at a fixed level (65 dB SPL) of noise while adjusting the sentence level adaptively based on the listener’s response. The signal-to-noise ratio (SNR) required to achieve 50% intelligibility (speech recognition thresholds, SRTs) was obtained. Results: On average, the aided SRT of HI listeners was 0.1 dB SNR, and the mean SRT of NH adults was -8.91 dB SNR. The Matrix sentence-in-noise intelligibility was not sufficiently explained by the unaided WRS or unaided SRS. Conclusion: A traditional measure of the unaided speech-in-quiet recognition cannot accurately predict the aided speech-innoise intelligibility. Clinically, a software-based sentence-in-noise intelligibility test is recommended to directly confirm the actual benefits of hearing aid in noisy situations.