Effects of frequency compression on the intelligibility and quality of speech in noise.

Abstract

Frequency-lowering technologies such as frequency compression have become common in high-end digital hearing aids. However, there are few data on the consequences of adjusting frequency compression ratio (FCR) and frequency compression cutoff (FCC). In this study, sentences at signal-to-noise ratios (SNRs) ranging from −10 to +10 dB were frequency compressed at a variety of FCRs (1.5–3) and FCCs (1–2 kHz). Frequency compression was implemented using sinusoidal modeling: the incoming signal was segmented, the spectral peaks found for each segment, and an output sinusoid at a shifted frequency was generated for each peak. Subjects were adult listeners with normal hearing or mild-moderate sensorineural loss. Intelligibility results illustrate a complex interaction between SNR and frequency compression parameters. At favorable SNRs, frequency compression had minimal effects. At less favorable SNRs, reducing FCC degraded intelligibility more than reducing FCR. At the poorest SNRs any amount of frequency compression degraded intelligibility but the specific FCC/FCR had little effect. Quality results show less interaction whereby any frequency compression was perceived as reducing speech quality. Compared to data from normal-hearing listeners, results from listeners with hearing loss were more variable and were interpretable in the context of audiogram configuration. [Work supported by NIH and GN ReSound.]