Related Experiment Video
Updated: May 21, 2026
09:29
A Standardized Obstacle Course for Assessment of Visual Function in Ultra Low Vision and Artificial Vision
Published on: February 11, 2014
Evaluating OCR performance for assistive technology: effects of walking speed, camera placement, and camera type
Junchi Feng1,2, Nikhil Ballem3, Mahya Beheshti3,4
1Department of Biomedical Engineering, Tandon School of Engineering, New York University, Brooklyn, NY, USA.
Disability and Rehabilitation. Assistive Technology
|May 19, 2026
Summary
Optical character recognition (OCR) accuracy decreases with faster walking speeds and wider viewing angles. For mobile navigation, OCR systems need dynamic evaluations, not just static ones, to ensure practical performance.
Area of Science:
- Assistive Technology
- Computer Vision
- Human-Computer Interaction
Background:
- Optical Character Recognition (OCR) is crucial for assistive technology for visually impaired individuals.
- Existing OCR evaluations often use static datasets, failing to represent real-world mobile usage challenges.
Purpose of the Study:
- To evaluate OCR performance under static and dynamic walking conditions relevant to mobile navigation.
- To assess the impact of varying parameters like speed, viewing angle, and device placement on OCR accuracy.
Main Methods:
- Tested OCR engines (Google Vision, PaddleOCR, EasyOCR, Tesseract) on smartphones and smart glasses.
- Varied static conditions (distance 1-7m, angle 0°-75°) and dynamic conditions (walking speed 0.8-1.8m/s).
- Compared head-mounted, shoulder-mounted, and handheld device positions, calculating character-level accuracy.
Main Results:
- OCR accuracy decreased with increased walking speed and wider viewing angles.
- Google Vision showed the highest accuracy; PaddleOCR was the top open-source option.
- The phone's main camera and shoulder-mounted placement performed best, though differences in body position were not significant.
Conclusions:
- OCR performance is influenced by the engine, camera, field of view, device placement, and user motion.
- Dynamic, mobility-relevant evaluations are essential for OCR systems used in navigation.
- Future OCR systems should balance coverage, accuracy, and practical deployment for real-world use.
