The Susceptibility of Speech to Noise

Ohio State University, Columbus OH

Investigators

Abstract

Project Summary/Abstract The current topic involves the deleterious effect of noise on speech understanding. This has been a topic of considerable interest for many decades, motivated in part by the fact that difficulty understanding speech when background noise is present represents the primary auditory complaint of individuals with hearing loss. We argue here that a critical barrier to progress involves an incomplete understanding of the basic effect. A long- standing assumption, which is deeply ingrained in the field, is that the impact of noise on speech intelligibility is the same in all frequency regions. Said differently, adding noise at a given signal-to-noise ratio (SNR) is assumed to be equally detrimental to a band of speech in the low frequencies as it is to a band of speech in the middle or high frequencies. There exist many examples of this assumption, and it has shaped our thinking for many decades. However, in sharp contrast to this assumption, we recently revealed that the susceptibility of speech to noise is not at all the same across frequencies. Instead, the SNR required to impact speech equally varies across the spectrum by as much as 13 dB (Yoho et al., 2018). The proposed study will comprehensively establish this concept of “noise susceptibility.” In Aim 1, we will confirm, refine, and extend the initial finding. Noise at various levels will be added to individual speech bands to modulate their contribution to intelligibility between two well-defined boundaries. The overall amount of noise required to affect the contribution of a band, together with the rate of change, will determine its noise susceptibility. This will be done for each of 20 ANSI critical bands, resulting in detailed noise-susceptibility functions across frequencies. In Aim 2, we will expand our understanding of noise susceptibility by using maskers that impact speech via different mechanisms (Stone et al., 2011, 2012; Stone & Moore, 2014). Notionally steady noise operates primarily via modulation masking. Pure-tone maskers allow an examination of energetic masking. Noises that operate via modulation masking and have additional slow acoustic modulations introduce masking release. Each of these maskers will be used to create noise-susceptibility functions representing differential susceptibility across the speech spectrum, and the shapes of these functions will be compared. The information gained will reveal the degree to which maskers that operate via different mechanisms impact speech in different frequencies. This work will further our fundamental understanding of speech perception in noisy environments by advancing the noise-susceptibility concept and clarifying the various mechanisms of speech masking. The future implications of this knowledge are many. They include incorporation into a refined SII (ANSI, R2024) and the creation of noise-reduction systems tailored to focus on frequency regions where speech is least robust to noise, thus improving speech understanding for the 30 million Americans with hearing loss (NIDCD, 2024). Overall, the increased fundamental understanding, the rectification of decades of incorrect assumption, and the many implications have the potential to be transformative.