摘 要：
随着全球青少年心理健康问题的日益的凸显，如何把音乐疗愈融入日常中小学课堂的音乐教育中已成为跨学科研究的重要方向。本研究立足于青少年音乐教育与音乐疗愈课堂的交叉场域，构建并验证了一套名为“育愈和鸣”的人工智能音乐创作系统。本文的讨论和研究首先从神经音乐学、多模态情绪识别理论发展心理学的视角，生动系统地阐述了音乐疗愈课堂的运行理论基础与作用实践机制；其次，本研究设计并实现了一类融合视觉情感计算与声学情感分析的多模态情绪感知引擎，另基于Transformer架构的可控音乐生成引擎，形成了“情绪识别—特征映射—自适应音乐生成”的闭环干预框架；在此基础上，提出了“双源并行、动态干预”的课堂设计理念，架构了“核心曲库—扩展曲库—AI生成语料库”三位一体的系统引导性曲库，构建了音乐特征与PAD情绪间维度的量化控制、映射；本研究的最后，通过包含了自我报告、生理测量与行为观察的多模态情绪测量实验，验证了此方案在捕捉青少年情绪状态方面的一致性和有效性。
关键词：音乐疗愈课堂；人工智能音乐创作；多模态情绪识别；神经音乐学；青少年心理健康；音乐教育
Abstract:
With the increasing prominence of mental health issues among adolescents globally, integrating music therapy into daily music education in primary and secondary classrooms has become an important direction in interdisciplinary research. This study is grounded in the intersection of adolescent music education and music therapy classrooms, constructing and validating an artificial intelligence music creation system named “Yuyu Harmony”. The discussion and research in this paper firstly elaborates on the theoretical basis and practical mechanisms of music therapy classrooms from the perspectives of neuromusicology, multimodal emotion recognition theory, and developmental psychology. Secondly, this study designs and implements a multimodal emotion perception engine that integrates visual affective computing and acoustic emotion analysis, and a controllable music generation engine based on the Transformer architecture, forming a closed-loop intervention framework of “emotion recognition - feature mapping - adaptive music generation”. On this basis, it proposes a classroom design concept of “dual-source parallel and dynamic intervention”, constructs a three-in-one systematic guidance music library of “core music library - extended music library - AI-generated corpus”, and establishes quantitative control and mapping between music features and PAD emotions. Finally, through multimodal emotion measurement experiments involving self-reporting, physiological measurements, and behavioral observations, this study verifies the consistency and effectiveness of this approach in capturing the emotional states of adolescents.
Keywords: Music therapy classroom; Artificial intelligence music creation; Multimodal emotion recognition; Neuromusicology; Adolescent mental health; Music education
--
