Electrical sounds: preprocessed music for cochlear implant users
Keywords
Loading...
Authors
Issue Date
2021-01-27
Language
en
Document type
Journal Title
Journal ISSN
Volume Title
Publisher
Title
ISSN
Volume
Issue
Startpage
Endpage
DOI
Abstract
Music perception for cochlear implant users is limited due to inaccurate
pitch perception. As music is embedded in the cognitive system of humans,
it is important to improve music perception for cochlear implant users. This
study investigates how a neural network can improve pitch perception by
amplifying the resonant frequencies in music. In the end, all frequencies were
ampli ed instead of only the resonant frequencies, due to time constraints.
The neural network started with a loss of 120,000 and ended with a loss of
1000, which indicates it learned features from the spectrogram of the music
snippets. The online questionnaire did not show a signi cant result for
the comparison between the original vocoded music and optimised vocoded
music. There was a high variability between participants. The model can
be improved by only amplifying the resonant frequencies, either implicitly
or explicitly. The model can also be combined with other preprocessing
strategies, such as F0mod and semitone mapping.
Description
Citation
Supervisor
Faculty
Faculteit der Sociale Wetenschappen