Phosphene representation using Triplet Pose estimation
With stimulation of the visual cortex, we can generate the perception of so-called phosphenes, which are points of light perceptible without any light entering the eye. Through the usage of these phosphenes we can partially restore the vision of blind people by showing images on their visual cortex. This phosphene vision, however, is limited compared to regular vision due to having lower resolution and no colour contrast. The limitations mean that the images we want to stimulate onto the visual cortex need to be pre-processed, this is done so we only express the most important information, making the information more understandable. For the human body, a representation in phosphene vision can be made using the locations of their joints to create a skeleton. I present a model which uses triplet representation, a way of representing people using more coordinates per joint than a regular representation which provides more information about the dimensions of the human body. Using the triplet model, I was able to get more information about the shape of the human body. However, this information does not necessarily translate well into phosphene vision due to noise and research into the usability for blind people remains to be tested.
Faculteit der Sociale Wetenschappen