Objective and efficient evaluation of piano performance remains difficult due to the complex interaction among motor control, acoustic characteristics, and artistic expression. To improve the accuracy of performance assessment and training feedback, this study proposes a multimodal acoustic data-driven framework for piano performance error identification and correction. High-precision audio acquisition devices and motion capture sensors are employed to synchronously collect acoustic signals and key-motion parameters, forming a multimodal performance dataset. Based on signal analysis and pattern recognition techniques, an automated error identification model is developed to detect deviations in pitch accuracy, rhythmic stability, and dynamic control. Personalized correction strategies are subsequently generated through visual feedback and adaptive training modules. Experimental evaluation involving learners at different skill levels demonstrates that the proposed system significantly improves technical performance indicators compared with conventional training methods. Furthermore, the analysis reveals strong correlations between performance error patterns and practice behaviors. The proposed framework provides an effective approach for intelligent music education and offers methodological references for acoustic signal processing, multimodal sensing, and pattern recognition applications.
Y. S. Zhang (Thu,) studied this question.