6 96 6. Gaze-contingent processing improves mobility performance In the second experiment the number of phosphenes was reduced to 500 phosphenes to measure in minimal-vision conditions. The phosphene counts were determined using pilot experiments and are chosen for their associated level of difficulty: with the current values, the experimental tasks were difficult, but could still be performed. The visual field coverage was purposely limited to a small area to reflect possible limitations to the area that can be used for the electrode implantation. Furthermore, we expected the effects of the different gaze conditions to be more pronounced when phosphenes cover only a relatively small field of view. Note that the employed phosphene simulation reflects the visual quality of a hypothetical future implant with more electrodes and a larger spread of phosphenes compared to existing temporary prostheses (e.g., seeFernández et al., 2021). This choice enabled us to validate whether gaze-contingent processing yields benefits even for optimistic future implant designs. The encoding of the phosphenes (i.e., the activation pattern) was based on scene simplification with surface normal estimation combined with edge detection, which, also based on pilot experiments, turned out to be an intuitive phosphene representation for summarizing the current visual environments. 6.2.4. Experimental conditions Our study compares three experimental conditions (seeFigure6.1). In the gaze-locked simulation condition, the phosphenes encode what is captured by the head-centered virtual camera and the phosphene locations are coupled to eye movements. Note that while eye-tracking is required to simulate gaze-locked phosphene vision, the study condition simulates the experience produced by a head-steered visual prosthesis without an eye-tracking system. The gaze-contingent simulation condition simulates the effect of including a compensatory eye-tracking system in a head-steered prosthesis. Here, the phosphenes encode a region of the visual environment that matches with the gaze direction, enabling the user to sample visual input with eye-movements as well as head movements. The third study condition, the gaze-ignored condition, resembles commonlyused phosphene simulations that ignore the effects of eye movements on phosphene perception. Note that this simulation condition is included as a control condition to measure the isolated effects of head-steered visual sampling, and it does not accurately reflect the perceptual experience of cortical visual prosthesis users. 6.2.5. Experiment 1: Obstacle Avoidance Virtual Environment The design of the virtual environment was based on a previous (real-world) mobility study (de Ruyter van Steveninck et al., 2022b), and featured a straight corridor with boxes that served as obstacles (seeFigure 6.2bandFigure6.4). The virtual corridor was 44 meters long and consisted two empty sections and a middle section that contained 19 sets of obstacles. The obstacles were spaced evenly along the length of the corridor, two meters apart. The obstacle sets consisted either of a single small box or a combination of two large boxes, that were arranged similar to a partial wall or door-frame, spanning the entire height of the corridor and blocking two-thirds of its width. The placement of the obstacles varied between trials, but always matched one of the three unique templates. In total, each of the three hallway variants with a unique obstacle placement was always visited three times by each participant (once for every condition).
RkJQdWJsaXNoZXIy MTk4NDMw