gms | German Medical Science

Cochlear implant (CI) users with a hearing aid (HA) on the contralateral side (bimodal CI users) are a growing group of patients. These patients often show speech intelligibility improvements when they use both devices compared to CI-only or HA-only hearing. This study investigates the effect of noise reduction algorithms on simulated bimodal CI users. Simulation of CI was accomplished using a pulsatile vocoder mimicking both the details of CI signal processing and physiologic details of electric hearing. Simulation of aided hearing loss (HL) was accomplished using a multichannel dynamic compressor and a frequency- and level-dependent attenuation that mimics impaired loudness perception. Normal hearing (NH) subjects listened to speech in noise in three simple anechoic situations through the CI simulation on the right and the aided impaired simulation on the left ear. Speech reception thresholds (SRTs) with the Oldenburg sentence test were assessed.

Three algorithms implemented on the open Master Hearing Aid were evaluated, a single channel noise reduction algorithm (SCNR), monaural adaptive differential microphones (ADMs), and a binaural minimum variance distortionless response (MVDR) beamformer. These algorithms were compared to unprocessed conditions. The results indicated that SCNR does not have a significant effect on SRT. On the other hand, ADM showed a maximum SRT-improvement of 11.3 dB (NH), 13.2 dB (mild HL), and 11.7 dB (severe HL). The MVDR implementation (although not optimal) showed a maximum SRT-improvement of 1.3 dB (NH), 11.3 dB (mild HL) and 8.3 dB (severe HL). Benefit of the ADM was mostly independent of simulated impairment, whereas the benefit of the MVDR depended on the simulated impairment and the tested signal-to-noise ratio (SNR). This indicates that a prediction of algorithm performance should take the SNR and a model of the impairment into account.