Scaling multimodal learning with TouchMotion AI: a design framework for embodied Thai dance instruction in multi-disability Blind education

Abstract

Traditional instruction of Thai classical dance relies heavily on contact-based physical guidance, which may limit opportunities for independent practice among visually impaired students with Multiple Compounded Disabilities (MCD). This study developed and implemented the TouchMotion AI Platform through a design-based, exploratory mixed-methods study focused on technical feasibility, user acceptance, and classroom usability rather than controlled educational efficacy. The platform combines edge-computed pose estimation, 3D tactile interfaces, and real-time corrective spatial audio feedback to support non-visual movement practice. It was implemented across four specialized schools in Northern Thailand with 80 students and 24 educators. An initial 20-case validation subset demonstrated high technical reliability. Educator-completed learning-practice assessments indicated favourable perceptions of learners’ ability to use the feedback during posture-adjustment practice (M = 4.45, SD = 0.59, on a 5-point scale). Student interviews and educator observations suggested reduced movement-related hesitation and repeated practice with less direct physical guidance. These findings represent exploratory implementation outcomes rather than independently verified educational effects. Overall, the platform provides a practical model for accessible embodied arts learning and illustrates how AI-supported tactile and auditory feedback may expand participation in culturally significant performing-arts activities.