Source detection: Influence of the Doerbecker’s noise reduction algorithm; Simulation in the case of binaural cochlear implant coding

Abstract

Cochlear implant (CI) provides good auditory performances in quiet, but recognition in noisy environments is still a challenge for CI users. Bilateral hearing can improve speech perception; it has been shown with normal hearing aids, but how source localization is affected is still an open question.In this study source localization is studied in the case of a binaural cochlear implant coding. Signal was received on the two ears and then processed using the classical noise reduction algorithm, suggested by Doerbecker and based on the spectral properties of the signal. Two correction strategies, Ephreim & Malah and Scalart, have been inserted in the processing chain.Tests were conducted, “in simulation” with 20 normal hearing listeners who received a cochlear implant coded signal. An analytic interpretation of the results offers a frame to quantify the different contributions.Results show that the Doerbecker’s algorithm affects the source localization. This effect was lowered when a percentage of the input signal was re-injected into the system. Both correction strategies were equivalent.