evidoria

← Back to browse

Good practice Imported

Real-Time Palestinian Sign Language Recognition Reaches 97.6% Accuracy for Deaf Students' Math Lessons

Palestine, State of · Jenin · See the Palestine, State of profile

Evidence: Descriptive / self-reported Top 70% 40/100 · Ask Evidence Copilot about this practice

Researchers at Palestine's Arab American University, backed by Warwick's Education Fund, built a Vision Transformer recognising Palestinian Sign Language gestures for 41 math concepts at 97.59% accuracy, giving deaf pupils real-time math lessons in their own sign language.

97.59 %
PSL math-gesture recognition accuracy
41
PSL gesture classes in dataset
434
Sign Pulse annotated classroom video clips

Details

Maturity
Pilot
Promoter
Arab American University (AAUP), Palestine, with University of Warwick Education Fund support
Period
2025 (Vision Transformer study published July 2025); related Sign Pulse framework published 2026
Keywords
assistive technology, computer vision, sign language recognition, mathematics education, deaf and hard-of-hearing accessibility

Context

Deaf and hard-of-hearing pupils in Palestine have had almost no dedicated digital resources in Palestinian Sign Language, leaving them dependent on interpreters or written Arabic for subjects like mathematics.

Activities

Researchers at the Arab American University, with University of Warwick Education Fund support, built a custom dataset of 41 PSL gesture classes representing core mathematical concepts and trained a Vision Transformer to recognise them in real time; a related Sign Pulse framework extends this across STEM subjects for grades 1-4 using a newly built 434-clip dataset.

Results

The recognition model reached 97.59% accuracy.

Conclusions

The recognition model reads hand gestures only, missing facial expressions and body movement that carry meaning in PSL, and no study yet measures whether pupils' actual math learning improves from using the system.

Implementation

Indicative cost
Low (< €50k)
Time to results
Short (< 1 year)

Do you run this practice? Claim it — verified implementers get a public contact pathway and can propose corrections.

Data sources

Where this practice's information was retrieved from, and when.

Similar practices you may find useful