2 18 2. Real-world indoor mobility with simulated prosthetic vision Resolution 10x10 18×18 26×26 34×34 42×42 50×50 Trial duration <0.001 <0.001 0.044 0.765 0.145 <0.001 No. of collisions 0.003 <0.001 0.004 0.433 0.889 0.345 Subjective rating 0.039 0.002 0.124 0.078 0.221 0.039 Table 2.4: p-values for Wilcoxon signed rank test for evaluation of the effect of scene complexity with Canny edge detection. With a Bonferroni correction of αfor six planned comparisons, findings are considered significant if p<0.0083. 18). Notably, in the higher phosphene resolutions this effect was absent or even opposite. With a resolution of 50 × 50 phosphenes, participants achieved a significantly lower trial duration in the complex environment compared to the simple environment. 2.3.4. The effect of image processing: SharpNet versus CED The p-values of the Wilcoxon signed-rank test on the effect of image processing are provided inTable2.5and the performance is visualized inFigure2.6. For the 42 × 42 phosphenes condition in the complex environment, SharpNet trials were significantly longer (p < 0.001) and received a lower subjective rating (p < 0.001) compared to CED. In the 26 × 26 phosphenes condition no significant performance differences were found in the complex environment. In the simple environment performance was worse for the SharpNet condition compared to CED regardless of the phosphenes resolution, as reflected by longer trial durations (p < 0.001 for 42 × 42 phosphenes, p = 0.005 for 26 × 26 phosphenes) and lower subjective rating (p < 0.001 for 42 × 42 phosphenes, and p = 0.004 for 26 × 26 phosphenes). Simple scene Complex scene Resolution 26×26 42×42 26×26 42×42 Trial duration 0.005 <0.001 0.478 <0.001 No. of collisions 0.055 0.084 0.331 0.169 Subjective rating 0.004 <0.001 0.520 <0.001 Table 2.5: p-values for Wilcoxon signed-rank test for evaluation of the effect of image processing method. With a Bonferroni correction of αfor four planned comparisons, findings are considered significant if p < 0.0125 2.3.5. User experience After the experiment, participants had the opportunity to indicate their personal experience in a survey. Here we report some relevant observations. 82.4% of the respondents indicated that they experienced a sufficient amount of practice during the practice session, versus a 17.6% who indicated that more practice trials would have been beneficial. Upon asking for specific cues that influenced their navigation strategy, 94.1% of the respondents answered that they used the object contours for recognizing obstacles and 47.1% of the respondents indicated that for some of the trials, choices in navigation were based on chance or intuition rather than understanding of the visual input. Furthermore, 53.0% of the respondents explicitly indicated to make use of apparent differences in lines and textures on the floor, boxes and walls. In the “other comments” section, some participants mentioned that the (in)ability to perceive depth strongly influenced the performance for
RkJQdWJsaXNoZXIy MTk4NDMw