Standardized Teaching Aid System for Piano Performance Movements Integrating Posture Estimation and Temporal Feature Extraction

Current piano performance training primarily relies on subjective observation and experience-based assessment, limiting the objective evaluation of posture accuracy and temporal coordination. This study proposes a standardized teaching aid system that integrates an improved posture estimation network with temporal feature extraction to achieve intelligent analysis and real-time feedback for piano performance movements. The system extracts upper-limb and finger keypoints through a lightweight posture estimation framework enhanced by spatial attention mechanisms and constructs spatiotemporal representations using a Temporal Transformer with embedded temporal position encoding. A quantitative evaluation network is further developed to assess spatial coordination, temporal consistency, and fingering stability, generating standardized scores and dynamic feedback signals in a closed-loop teaching framework. Experimental results demonstrate a keypoint localization stability rate of 96.0%, feedback accuracy of 97.2%, and response delays ranging from 0.48 s to 0.77 s across different performance scenarios. The proposed framework exhibits strong robustness and adaptability for complex motion analysis. It is particularly applicable to intelligent sensing environments supported by wireless communication infrastructures and antenna-enabled edge devices, where reliable real-time transmission of visual and motion data is essential for responsive feedback and human–machine interaction. This study provides an effective engineering solution for intelligent motion assessment, standardized skill training, and data-driven interactive learning systems.