AI-Based Speech Therapy Assistant for Malayalam: A Deep Learning Approach for Language-Specific Speech Disorders

Speech therapy is a clinical intervention that addresses communication disorders affecting speech production, language comprehension, fluency, and voice quality. While essential for improving quality of life and social participation, speech therapy faces significant challenges including professional shortages, accessibility barriers, engagement difficulties, subjective assessment practices, and limited personalization. These challenges are magnified for low-resource languages like Malayalam, which lacks standardized assessment tools, therapy materials, and trained professionals familiar with its unique phonological features such as distinctive phonemes and geminate consonants. Conventional speech therapy approaches designed for widely spoken languages often fail to address language-specific articulation patterns, creating a critical need for technology-assisted solutions that can overcome geographical barriers, standardize assessment procedures, and preserve linguistic authenticity while providing accessible intervention for Malayalam speakers with speech disorders. This paper proposes an AI-driven speech therapy assistant tailored for Malayalam speakers with speech disorders. The proposed system leverages deep learning techniques to address common articulation errors, such as mispronunciations and fluency issues like stammering. The framework aims to analyse speech patterns, identify errors, and recommend personalized therapeutic exercises. By focusing on Malayalam-specific phonetics, this research bridges the gap in accessible and culturally relevant speech therapy solutions.