Recognition of objects in simulated irregular phosphene maps for an epiretinal prosthesis

Area of Science:

• Biomedical Engineering
• Computer Vision
• Neuroscience

Background:

• Restoring vision to the blind using visual prostheses is a significant challenge.
• Understanding the minimum visual requirements for daily tasks is crucial for effective prosthesis design.

Purpose of the Study:

• To investigate the impact of simulated phosphene map parameters on object recognition.
• To determine the minimum requirements for visual prostheses to enable recognition of common objects.

Main Methods:

• Simulated irregular phosphene maps were used to assess object recognition.
• The effects of resolution, distortion, dropout percentage, and grayscale were systematically evaluated.
• Two image processing methods were compared under high distortion or dropout conditions.

Main Results:

• Object recognition accuracy significantly increased with higher resolution.
• Increased distortion and dropout percentage led to a considerable decrease in recognition accuracy.
• Grayscale had a significant impact only at level 2; other levels showed minimal difference.

Conclusions:

• Resolution is a key factor for effective object recognition in visual prostheses.
• Minimizing distortion and dropout is critical for improving visual prosthesis performance.
• Image processing techniques can mitigate performance degradation under challenging visual conditions.