Optimizing Stimulation Strategies in Cochlear Implants for Music Listening

Most cochlear implant (CI) strategies are optimized for speech characteristics while music enjoyment is signicantly below normal hearing performance. In this thesis, electrical stimulation strategies in CIs are analyzed for music input. A simulation chain consisting of two parallel paths, simulating normal hearing conditions and electrical hearing respectively, is utilized. One thesis objective is to congure and develop the sound processor of the CI chain to analyze dierent compression- and channel selection strategies to optimally capture the characteristics of music signals. A new set of knee points (KPs) for the compression function are investigated together with clustering of frequency bands. The N-of-M electrode selection strategy models the eect of a psychoacoustic masking threshold. In order to evaluate the performance of the CI model, the normal hearing model is considered a true reference. Similarity among the resulting neurograms of respective model are measured using the image analysis method Neurogram Similarity Index Measure (NSIM). The validation and resolution of NSIM is another objective of the thesis. Results indicate that NSIM is sensitive to no-activity regions in the neurograms and has diculties capturing small CI changes, i.e. compression settings. Further verication of the model setup is suggested together with investigating an alternative optimal electric hearing reference and/or objective similarity measure