| Literature DB >> 21927522 |
Abstract
Frequency lowering is a form of signal processing designed to deliver high-frequency speech cues to the residual hearing region of a listener with a high-frequency hearing loss. While this processing technique has been shown to improve the intelligibility of fricative and affricate consonants, perception of place of articulation has remained a challenge for hearing-impaired listeners, especially when the bandwidth of the speech signal is reduced during the frequency-lowering processing. This paper describes a modified vocoder-based frequency-lowering system similar to one reported by Posen, Reed, and Braida (1993), with the goal of improving place-of-articulation perception by enhancing the spectral differences of fricative consonants. In this system, frequency lowering is conditional; it suppresses the processing whenever the high-frequency portion (>400 Hz) of the speech signal is a periodic signal. In addition, the system separates non-sonorant consonants into three classes based on the spectral information (slope and peak location) of fricative consonants. Results from a group of normal-hearing listeners with our modified system show improved perception of frication and affrication features, as well as place-of-articulation distinction, without degrading the perception of nasals and semivowels compared to low-pass filtering and Posen et al.'s system.Entities:
Year: 2012 PMID: 21927522 PMCID: PMC3171973 DOI: 10.1016/j.specom.2011.07.008
Source DB: PubMed Journal: Speech Commun ISSN: 0167-6393 Impact factor: 2.017