Signal processing in quiet and noise

Abstract

It has been shown that many profoundly deaf patients using multichannel cochlear implants are able to understand significant amounts of conversational speech using the prosthesis without the aid of lipreading. These results are usually obtained under ideal acoustic conditions but, unfortunately, the environments in which the prostheses are most often used are rarely perfect. Some form of competing signal is always present in the urban setting, from other conversations, radio and television, appliances, traffic noise and so on. As might be expected, implant users in general find background noise to be the largest detrimental factor in their understanding of speech, both with and without the aid of lipreading. Recently, some assessment of implant patient performance with competing noise has been attempted using a four-alternative forced-choice spondee test (1) at Iowa University. Similar testing has been carried out at the University of Melbourne with a group of patients using the Nucleus multichannel cochlear prosthesis. This study formed part of an assessment of a two formant (F0/FI/F2) speech coding strategy (2). Results suggested that the new scheme provided improved speech recognition both in quiet and with competing noise. This paper reports on some more detailed investigations into the effects of background noise on speech recognition for multichannel cochlear implant users.