Design of an intelligent aesthetic education teaching system based on multimodal signal interaction

The application of intelligent algorithms in educational settings has driven the structured and intelligent development of aesthetic education. Addressing the heterogeneous characteristics of speech, visual, and physiological signals in multimodal data, this study constructs an intelligent aesthetic education teaching system integrating perception, feature extraction, and feedback generation. The system employs a joint modeling framework integrating CNN, BLSTM, and Transformer architectures, incorporating cross-modal attention mechanisms to achieve dynamic signal weighting and feature collaboration. After training on 26,800 multimodal samples, it achieves 94.8% accuracy in emotion recognition with an average response latency of 230ms. Experimental results demonstrate the system's high real-time performance and stability, providing computational support and an extensible framework for intelligent teaching feedback.