Help us understand how this website is used. You can choose essential storage only. Privacy information
MIXED REALITY / AR IN EDUCATION
GeoLex AR explores a common interaction architecture across geometry and foundational literacy, while preserving specialized processing for each learning task.
THE RESEARCH, EXPLAINED
GeoLex AR brings geometry and English literacy into one Android learning suite. Its central idea is to respond to something the child creates: a net, a fingertip movement or a sketch. Each activity uses its own processing method before returning feedback or a 3D/AR learning experience.
C1 / GEOMETRY VALIDATION
A child constructs a flat arrangement of faces. Shape-specific mathematical rules check whether it can form a supported solid. Valid nets progress to 3D and AR; invalid ones move to corrective guidance.
C2 / CORRECTION & SLICING
Visual guidance highlights a faulty relationship in the net so the child can revise it. For valid solids, guided slicing activities reveal prepared sections and fractions such as halves and quarters.
C3 / AIR-WRITING
The mobile camera follows the child’s fingertip in the air. The ordered movement is processed by a trained recognition model to return a predicted letter and feedback.
C4 / SKETCH & VOCABULARY
The child draws an apple, cat, fish, sun or tree. A classifier identifies the sketch; spelling and pronunciation-oriented activities connect it to language. Completing both activities unlocks its AR representation.
The pilot involved 50 unique learners: 25 aged 9–12 tried the geometry components, and 25 aged 6–8 tried the literacy components. The reported results describe this pilot; they do not establish effectiveness for every learner or classroom.
Explore the pilot findings →LITERATURE SURVEY & RESEARCH GAP
AR can make learning content inspectable. Its educational value depends on the interaction design, the task and the learner’s age.
The manuscript positions AR geometry studies as evidence for spatial learning, while noting their usual focus on predefined objects. Sketch recognition and reconstruction interpret freehand input, but do not by themselves establish a net’s mathematical validity or provide educational correction.
Camera-based air-writing preserves ordered fingertip movement. AR language-learning research supports interactive alphabet and vocabulary work. GeoLex AR investigates how computational interpretation can sit between the learner’s own input and reinforcement across both pathways.
How can an integrated mobile suite validate or recognize learner-generated input, return appropriate instructional feedback and support age-appropriate 3D/AR reinforcement?
Positioning follows the manuscript, pp. 1–2. The contribution is architectural and pedagogical; novelty is not asserted merely because four features appear in one app.
MAIN OBJECTIVE
Learners construct a geometric net. Shape-specific rules check its mathematical validity before a solid appears in 3D and AR.
Fault-specific guidance helps explain an invalid net. Once a solid is ready, virtual slicing makes its internal structure and fractions visible.
The mobile camera tracks a fingertip moving through free space. An ordered trajectory becomes the input to CNN-based letter recognition.
A learner’s sketch becomes an object-word connection. Spelling and pronunciation-oriented activities unlock its corresponding AR representation.
DESIGN SCIENCE RESEARCH
The study uses Design Science Research: the mobile suite is an artifact iteratively designed and evaluated. Technical workflow evidence is separated from learner outcomes.
50 unique learners participated: the same 25 aged 9–12 evaluated C1/C2; the same 25 aged 6–8 evaluated C3/C4. The findings are descriptive.
Manuscript pp. 2–3. Evaluation detail →
SYSTEM ARCHITECTURE
↺ Revise net → retry C1 validation
Accessible architecture reconstruction based on manuscript pp. 3–4, Figures 2–3. The correction loop returns invalid geometry to validation. The literacy branches remain separate.
TECHNOLOGIES USED
Unity 2022.3 LTS, C# and Android. AR Foundation with the ARCore XR Plugin supplies plane detection, raycasting, placement and physical-viewpoint inspection.
C1 uses deterministic geometry rules. C3 uses camera hand-landmark tracking and CNN-based trajectory recognition. C4 development and integration use Python, TensorFlow/Keras, OpenCV, FastAPI, NumPy and Pillow. Earlier proposed technologies are not assumed to be final implementation choices.
Static Astro pages with TypeScript, CSS and small JavaScript interactions. The inquiry service uses Firebase Cloud Functions, private Cloud Firestore storage and server-side SMTP.
Research technology claims: manuscript p. 3.
SELECTED RELATED WORK
Nadzeri et al. (2024), International Journal of Interactive Mobile Technologies. Manuscript reference [3].
Uriarte-Portillo et al. (2023), Information. Manuscript reference [6].
Watanabe et al. (2023), Electronics. Manuscript reference [14].
Korosidou (2024), Digital. Manuscript reference [16].
Weigand, Johannesson & Andersson (2021), Data & Knowledge Engineering. Manuscript reference [19].
These references are transcribed from the supplied manuscript. Read the complete references on page 6.
Open the manuscript bibliography ↗