Optimization of Sound Coding Strategies to Make Singing Music More Accessible for Cochlear Implant Users

Cochlear implants (CIs) are implantable medical devices that can partially restore hearing to people suffering from profound sensorineural hearing loss. While these devices provide good speech understanding in quiet, many CI users face difficulties when listening to music. Reasons include poor spatial specificity of electric stimulation, limited transmission of spectral and temporal fine structure of acoustic signals, and restrictions in the dynamic range that can be conveyed via electric stimulation of the auditory nerve. The coding strategies currently used in CIs are typically designed for speech rather than music. This work investigates the optimization of CI coding strategies to make singing music more accessible to CI users. The aim is to reduce the spectral complexity of music by selecting fewer bands for stimulation, attenuating the background instruments by strengthening a noise reduction algorithm, and optimizing the electric dynamic range through a back-end compressor. The optimizations were evaluated through both objective and perceptual measures of speech understanding and melody identification of singing voice with and without background instruments, as well as music appreciation questionnaires. Consistent with the objective measures, results gathered from the perceptual evaluations indicated that reducing the number of selected bands and optimizing the electric dynamic range significantly improved speech understanding in music. Moreover, results obtained from questionnaires show that the new music back-end compressor significantly improved music enjoyment. These results have potential as a new CI program for improved singing music perception.