The just-noticeable difference in speech-to-noise ratio.

Abstract

Just-noticeable differences (JNDs) have been measured for various features of sounds, but despite its importance to communication, there is no benchmark for what is a just-noticeable—and possibly meaningful—difference in speech-to-noise ratio (SNR). SNR plays a crucial role in speech communication for normal-hearing and hearing-impaired listeners. Difficulty hearing speech in background noise—a poor SNR—often leads to dissatisfaction with hearing-assistance devices. While such devices attempt through various strategies to address this problem, it is not currently known how much improvement in SNR is needed to provide a noticeable benefit. To investigate what is a noticeable benefit, we measured the JND in SNR for both normal-hearing and hearing-impaired listeners. Here, we report the SNR JNDs of 69 participants of varying hearing ability, estimated using either an adaptive or fixed-level procedure. The task was to judge which of the two intervals containing a sentence in speech-spectrum noise presented over headphones was clearer. The level of each interval was roved to reduce the influence of absolute level cues. The results of both procedures showed an average SNR JND of 3 dB that was independent of hearing ability. Further experiments using a subset of normal-hearing listeners showed that level roving does elevate threshold. These results suggest that noise reduction schemes may need to achieve a benefit greater than 3 dB to be reliably discriminable.