An AI-Assisted Educational Framework for Physical Skill Acquisition via Inertial Motion Sensing and Generative Learning Models

Abstract

Artificial intelligence (AI) has transformed cognitive education, yet its integration into physical-skill learning remains limited. Traditional sports training still relies on in-person coaching, costly equipment, or subjective self-assessment, restricting accessibility. This paper presents an AI-assisted educational framework that combines low-cost wearable inertial sensing with a generative language model (GLM) to deliver personalized, realtime coaching. As a case study, the framework was implemented for cue sports using an ESP32-based inertial measurement unit (IMU) connected to a mobile app with visualization, history tracking, and a GPT-series coaching assistant. The system employs a hybrid feedback strategy: a rule-based motion-scoring layer for objective consistency metrics and a GLM layer that converts these metrics and shot context into natural-language guidance. Evaluation with 25 amateur players indicated high usability (System Usability Scale, SUS: 91/100), sub-2 s latency, and over 90 % perceived feedback accuracy. In addition to player self-ratings, an experienced coach qualitatively reviewed a subset of sessions, providing two-way verification of plausibility. While dataset size and connectivity remain limitations, results confirm the feasibility of using GLMs to augment physical skill acquisition. Beyond cue sports, the framework generalizes to domains such as tennis, golf, or rehabilitation, aligning with the goals of smart education.

Ähnliche Arbeiten