Huruwa: An AI-IoT Robotic System for Adaptive Speech Therapy and Parental Support for Sinhala-Speaking Children with Speech Sound Disorders

Abstract

Early phonological intervention for Sinhalaspeaking children with speech sound disorders requires scalable, engaging tools that bridge clinical expertise and home practice; however, low-resource language constraints limit available solutions. This paper introduces Huruwa, an AI-IoT robotic platform designed to support speech therapy for Sinhala-speaking children. The system integrates LLM-guided conversational interaction, phoneme-adaptive exercises (80-90% task suitability), and SVM-RBF-based phoneme error detection achieving 78% accuracy. It further employs knowledge-graph-driven therapy generation and a RAG-based parent guidance system to deliver grounded, hallucination-controlled support. Evaluations confirm real-time feasibility across components, offering a deployable model for Sinhala child speech therapy in resource-limited settings like Sri Lanka.