An Improved Processing Strategy for Head Pose Detection in Phosphene Vision
Keywords
Loading...
Authors
Issue Date
2021-01-29
Language
en
Document type
Journal Title
Journal ISSN
Volume Title
Publisher
Title
ISSN
Volume
Issue
Startpage
Endpage
DOI
Abstract
Cortical visual implants can elicit activity in the visual cortex via electrical impulses.
Speci c stimulation patterns can partially restore vision in the form of
phosphene images. However, because of the limited resolution of these images only
a small part of the visual eld can be e ectively represented. Intelligent processing
strategies can help with transforming the visual eld image into a meaningful
phosphene representation. I propose a new processing strategy for head pose position
in phosphene vision using the vertices of a 3D model and call it \vertex processing".
Multiple variations of this processing strategy were implemented and compared to
each other as well as to an edge detection algorithm, which is a commonly used processing
strategy for cortical visual implants at the moment. Accuracy and reaction
time data from participants in online experiments was compared. I found that there
was a signi cant improvement in these metrics for the vertex processing strategy compared
to the current edge detection strategy. The head pose directions that pro ted
most from this new strategy were up, down, front and down right. I also found that
the amount of vertices in the 3D model could be reduced to a certain extent without
signi cant accuracy loss, but not without a signi cant increase in reaction time.
These results indicate that the vertex processing strategy can create a more e ective
representation of the visual eld and in this way can contribute to a more meaningful
visual experience for cortical visual implant users.
Description
Citation
Supervisor
Faculty
Faculteit der Sociale Wetenschappen