Augmented Reality–Based English Learning Media Using a Target Tracking Model Approach

Abstract

Young learners often find English vocabulary hard because spelling differs from pronunciation, and teaching by explanation alone is monotonous. This study built BINGDAR, an Android application in which pointing a phone at a real object (fruit, classroom object, or animal picture cube) shows a rotatable, zoomable 3D model with audio and text, using Vuforia model targets in Unity, and evaluated it with 30 third-grade students. All 15 interface functions passed the black-box test. In the target-detection test, the models were recognized between 30 and 60 cm for all three categories but never at 5 cm, and the viewing angle mattered in only 2 of 45 distance–angle conditions. Mean scores rose within each of five pretest–posttest sessions (45→64, 65→75, 81→89, 94→96, 97→98); the shrinking percentage gains (42% to 1%) mostly reflect the ceiling, since the normalized gain stayed between 0.29 and 0.42. Because there was no control group, these gains cannot be attributed to the application alone. The mean System Usability Scale score was 91.3 (95% CI 89.8–92.8), and all six User Experience Questionnaire scales (2.03–2.44) exceeded the “excellent” benchmark even at the lower confidence limit, although perspicuity had low reliability (α = 0.50). The application is well accepted; evidence for its learning effect needs a comparison group and individual scores.