Minimum Requirements for Mobility in Unpredictible Environments

Abstract

Purpose: : Determination of the minimum requirements for a visual prosthesis to restore useful mobility in random environments with dynamic elements.

Methods: : Normal volunteers performed 2 mobility tasks under conditions simulating artificial vision: walking through a random maze of obstacles (random forest – 6 subjects) and estimating of the danger in an every–day traffic situation (road crossing – 4 subjects). Various degrees of image resolution (17920 down to 124 pixels) and several effective visual fields (8.3°x6° up to 66°x46°) were projected on a 10°x7° viewing area, stabilized on the central retina. Performance was assessed using objective measures (e.g. number of errors or time to accomplish the task) as well as subjective measures (questions to the subject).

Results: : Random forest task: Only a slight and non–significant tendency toward higher error counts could be observed in difficult situations. Subjects essentially compensated the lack of visual information with longer times to accomplish the task. Time per run decreased significantly (p<0.05) for image resolutions of 498 pixels and below. Errors demonstrated a slight but non–significant tendency to increase at low pixel numbers. While there was no significant difference between the different effective visual fields, 33°x23° tended clearly to be the best compromise and was preferred by the subjects. Road crossing task: Subjects mainly compensated the lack of visual information with audition. They almost never ‘crossed’ the road at moments of danger. Decision time was mainly dependent on the actual traffic density. Subjective ‘difficulty’ and ‘security’ measures indicate, however, that image resolutions of more than 1000 pixels were highly appreciated for taking ‘safe’ decisions.

Conclusions: : Our results demonstrate that mobility in random environments including dynamic elements needs more visual information than mobility in well–known static environments. We estimate that about 500 stimulation points distributed over a 3x2 mm² retinal area are enough to purely accomplish almost any mobility task. An image resolution of 1000–2000 pixels would, however, greatly improve the feeling of security.

Keywords: space and scene perception • low vision • retina