DyslexiaMate: A Mobile Based System for Supporting Individual with Dyslexia Using Machine Learning Algorithms

Document Type

Conference Proceeding

Publication Date

2026

Abstract

Dyslexia is a common learning difficulty that significantly affects reading, writing, and phonetic recognition, particularly among Arabic-speaking learners, where existing solutions often lack linguistic adaptation and interactive support. This paper presents DyslexiaMate, a mobile-based system that combines phonetic-based learning techniques, assistive tools such as text-to-speech and speech-to-text, gamification elements, and object detection to enhance Arabic literacy skills for dyslexic individuals. The front-end interface is designed for accessibility, featuring calming splash and onboarding screens, a centralized home screen, and integrated tools like a speech-to-text converter and a “Forgotten Letters” game to reinforce recognition through playful engagement. The Maqroo font is used throughout the interface to enhance readability for dyslexic users. The back-end leverages machine learning models trained on the Phonics Exercise Audio Dataset and applies deep learning techniques for Arabic handwritten character classification. A YOLOv8s-based object detection module enables real-time recognition of Arabic letters using a mobile camera, supporting multisensory interaction. Experimental results demonstrate promising performance, with machine learning models achieving 92.8% accuracy on the Phonics Exercise Audio Dataset and deep learning models achieving 91% accuracy on the Arabic Handwritten Characters Dataset, confirming DyslexiaMate’s effectiveness in supporting early literacy interventions for Arabic-speaking individuals with dyslexia.

Link to Full Text

Share

COinS