The influence of onset enhancement on speech intelligibility in cochlear implants - A model based and CI simulation evaluation

Speech understanding in cochlear implants (CI) is good in quiet, but very poor in more adverse listening conditions. One of the reasons that normal hearing listeners have a better representation of the auditory scene in noisy environment than CI recipients, is the higher firing rate on the auditory nerve at onsets of signals. This is caused by the rapid adaptation effect that is present at the auditory nerve synapse. The better representation of the onsets leads to a better coding of the acoustical landmarks. The rapid adaptation effect is not present in CI stimulation, because the CI bypasses the synapse. An approach to improve speech understanding in noisy environment is mimicking the rapid adaptation effect of the auditory nerve synapse by emphasizing the onsets of the speech signal. Two CI processing strategies, enhanced envelope continuous interleaved sampling (EECIS) and transient emphasis spectral maxima (TESM), have been suggested before that include the mimicking of the adaptation effect. Both strategies show benefits in quiet, but fail in adverse listening conditions, when the access to acoustical landmarks seems to be most important. In this study, we investigate the influence of onset enhancement by comparing the EECIS, a variation of the TESM that was optimized to maximum gain at onsets, and the continuous interleaved sampling (CIS) as the reference strategy in two manners. The first manner, model based evaluation of CI strategies, compares the strategies by simulating the electrically evoked compound action potential (ECAP) response on the auditory nerve based on a model (collaboration with Laboratory of Experimental Audiology (LEA), Zurich). The second manner, evaluation with CI vocoder simulations, investigates the effect of onset enhancement in noisy environment by conducting subjective listening tests with vocoded speech with normal hearing listeners. A loudness rating experiment was performed to ensure that the influence of the enhanced onsets instead of the use of loudness cues is measured. The effect of the onset enhancement was present in the simulated responses of the ECAP model and differences between the strategies could be simulated. For the subjective listening tests, significant effects were observed for the strategies in comparison to the unprocessed signal under ideal conditions for the extraction of the enhanced signal. A speech reception threshold (SRT) improvement of 2.5 dB for the EECIS and 1.5 dB for the TESM strategy was obtained. The results of another series of subjective listening tests show that the EECIS strategy is applicable in noisy environment by introducing a noise reduction step and benefits in speech intelligibility were observed. In the loudness rating experiment, no significant effects between the strategies were obtained. The results of the evaluation based on an ECAP model show the potential benefits of the evaluation of CI strategies using simulations. The subjective listening test results demonstrate the importance of the onsets in speech and their contribution to speech intelligibility in noisy environment. Subjective listening tests have to be done in future to investigate the correlation between the simulated responses and the perception of onsets.status: publishe