Just-noticeable-difference noise, not map settings, drives vowel-identification variability between cochlear implant users

Evidence

The MPI model has been used to model the performance of the SCE strategy for five virtual CI users. All virtual CI user differs from each other in their most comfortable and threshold levels used in the map. That means that the speech processor of each virtual CI user will generate different electrodograms for the same vowels. Next, formant features were extracted from the electrodograms based on the spectral contrast between formants 1 and 2. Noise was added to the spectral contrast features (jnd noise). Three amounts of noise were added 0.01, 10 and 50 % of the magnitude of the formants extracted. … From the modelling results it can be observed that maximum performance is achieved for SCE factors 2 and 4. The other interesting aspect is that there is no difference in performance across the five different virtual CI users for “Noise Factor” 0.001 (Fig. 2a ). That means that the feature extraction (i.e. the extraction of the formants) is robust to the different electrodograms of each virtual CI user. Differences in performance between the five virtual CI users can only be observed for “Noise Factor” 0.1 and 0.5, meaning that the “jnd noise” is the only parameter explaining the differences. — Dijk, Pim van, Physiology, Psychoacoustics and…, 1858

From the modelling results it can be observed that maximum performance is achieved for SCE factors 2 and 4. The other interesting aspect is that there is no difference in performance across the five different virtual CI users for “Noise Factor” 0.001 (Fig. 2a ). That means that the feature extraction (i.e. the extraction of the formants) is robust to the different electrodograms of each virtual CI user. Differences in performance between the five virtual CI users can only be observed for “Noise Factor” 0.1 and 0.5, meaning that the “jnd noise” is the only parameter explaining the differences. — Dijk, Pim van, Physiology, Psychoacoustics and…, 1864

Categorization

• DEFINE: Clarifies that ‘jnd noise’ in the MPI model is the internal perceptual noise on spectral-contrast features, functionally distinct from the user’s threshold/map electrodogram differences.
• corroborated: 4 independent passages support this claim — Dijk, Pim van, Physiology, Psychoacoustics and…, 1858; Dijk, Pim van, Physiology, Psychoacoustics and…, 1864