Adaptive Mobile App Teaches Mexican Sign Language — TecNM/UABC Researchers Report 95.7% Gesture Accuracy and Strong Learning Gains

Mexican Sign Language (MSL) is the primary language of Mexico's deaf community, but its place in formal education has long been limited by a shortage of qualified interpreters and a lack of technology supporting personalised instruction. A research team from the Tecnológico Nacional de México (TecNM), based at the Instituto Tecnológico de Tijuana, began addressing this gap in 2023 under an approved TecNM research grant, jointly with the Universidad Autónoma de Baja California (UABC).

The project aimed to build a mobile app combining computer-vision gesture recognition with adaptive lesson difficulty to teach MSL, following a Scrum design process with input from educators, linguists and members of the deaf community.

The team built a gesture-recognition model using convolutional neural networks and pose-estimation, compressed for on-device use, and ran a controlled comparison of the adaptive app against a non-adaptive control group, published in peer-reviewed testing in Future Internet (MDPI) in September 2026.

The gesture-recognition model reached 95.7% top-1 accuracy (98.61% top-3) after compression from 38MB to 10.4MB, running at roughly 112ms per frame. Learners using the adaptive version achieved a normalised learning gain of 0.78 (48.2% to 88.7% sign-recognition accuracy), 84.3% retention on a delayed post-test, and 73.5% transfer-task accuracy versus 57.2% for the non-adaptive control group; usability scored 89.1/100 (System Usability Scale) with a Net Promoter Score of +72, and self-reported cognitive load (NASA-TLX) fell from 62 to 44.

These results come from a single peer-reviewed study by the developing team rather than an independently replicated, multi-site deployment, and the publication does not report the number of deaf students who took part or any classroom-scale rollout beyond the pilot.