Mobility and low contrast trip hazard avoidance using augmented depth.

Area of Science:

• Biomedical Engineering
• Computer Vision
• Neuroscience

Background:

• Current prosthetic vision systems often struggle with obstacle detection, particularly low-contrast ground obstacles.
• Existing visual representations for prosthetic vision lack sufficient emphasis on crucial environmental cues like ground-surface boundaries.

Purpose of the Study:

• To evaluate a novel visual representation called augmented depth for prosthetic vision.
• To determine if augmented depth improves mobility and reduces collisions compared to standard visual representations.

Main Methods:

• Human mobility trials were conducted with eight participants using simulated prosthetic vision.
• Participants navigated a course with low-contrast obstacles under four visual conditions: augmented depth, intensity-based, depth-based, and random.
• The novel augmented depth representation uses a ground plane extraction algorithm to increase contrast between obstacles and the ground.

Main Results:

• Augmented depth significantly reduced collisions by 48% compared to intensity-based, 44% compared to depth-based, and 72% compared to random representations.
• This indicates a substantial improvement in obstacle avoidance capabilities.

Conclusions:

• Augmented depth shows promise for enabling safer mobility for individuals with current and near-term prosthetic vision implants.
• This study demonstrates that enhancing visual scene augmentation to ensure key object visibility can lead to better prosthetic vision outcomes.