Related Experiment Videos
Optimisation of a computer vision system for the interpretation of American Sign Language
M H Abdallah1, A E Marble, C Charayaphan
1Department of Electrical Engineering, Technical University of Nova Scotia, Halifax, Canada.
Medical & Biological Engineering & Computing
|September 1, 1993
Summary
Optimizing camera angles for American Sign Language (ASL) interpretation is crucial. Positioning the camera within a specific 30-degree solid angle ensures 96% accuracy in recognizing ASL signs.
Area of Science:
- Computer Vision
- Human-Computer Interaction
- Linguistics
Background:
- Accurate interpretation of American Sign Language (ASL) is vital for communication accessibility.
- Current methods for ASLing recognition may be limited by camera positioning and depth information.
- Optimizing the viewing angle is essential for robust sign language recognition systems.
Purpose of the Study:
- To develop and evaluate a simulation algorithm for optimizing camera position for ASL interpretation.
- To assess the impact of depth information loss on ASL recognition accuracy.
- To determine the optimal viewing area for reliable ASL sign capture.
Main Methods:
- Simulation of a 3D world to 2D image point transformation.
- Analysis of depth information loss effects.
- Implementation of a sign projection correction test.
- Testing with two native/proficient ASL signers.
Main Results:
- A specific viewing area, subtending a 30-degree solid angle (45-degree azimuth, 45-degree elevation), is identified as optimal.
- Camera positioning within this area yields 96% accuracy for 36 tested ASL signs.
- Signer variability showed minimal impact on interpretation accuracy.
- Computer vision systems can differentiate visually similar signs using optimal depth information.
Conclusions:
- The defined camera positioning strategy significantly enhances ASL interpretation accuracy.
- The proposed method is robust to signer variability.
- Advanced computer vision techniques can overcome limitations of human visual perception in sign recognition.