Table of Contents

Abstract

Single-microphone noise reduction (SMNR) is commonly implemented in hearing aids to suppress background noise. Gain reduction by SMNR is usually applied to a mixture of speech and noise signals. Frication noise in fricative and affricate consonants mimics the randomness of broadband noise (e.g., white noise and pink noise) and the presence of fricative and affricate consonants within speech may be susceptible to gain reduction from the SMNR processing when noise is present. Essentially, this two-phase study was conducted to examine the acoustic and perceptual effects of SMNR on fricatives and affricates. In Phase 1, speech-plus-noise signals were presented to and recorded from each of the two hearing aids mounted on a manikin, under SMNR-on and SMNR-off conditions. Subsequently, the hearing-aid recordings underwent acoustic analyses to determine the effective signal-to-noise ratio. In Phase 2, recordings of Mandarin retroflex fricative and affricates, processed with HA#1 under SMNR-on and SMNR-off conditions, were used to examine the effects of SMNR on novel speech sound identification in noise by non-Mandarin speaking adults. Thirty-two self-reported non-Mandarin speakers with normal hearing participated in Phase 2. Phase 1 showed that SMNR processing in both hearing aids improved the output SNR of speech-plus-noise signals. Phase 2 showed that SMNR did not degrade the learning of novel speech sound identification among non-Mandarin speakers. In conclusion, although SMNR examined in this study caused acoustic changes in the amplitude domain, it did not impede the short-term learning of novel speech sound identification in background noise by non-Mandarin speakers.

Keywords: Hearing aid, microphone, digital processing, noise reduction, Mandarin