摘要
用户的情绪状态不同,需要的背景音乐也不同,因此提出基于LDA-MURE模型的背景音乐自适应推荐方法。提取背景音乐的音频特征和社会化标签,通过Fisher线性判别分析方法融合上述数据的特征,结合投影变换方法获得不同类别背景音乐的类内离散度和类间离散度。通过现代心理学分析人类情绪的节律周期变化,在此基础上判断用户当前的情绪状态。最后在LDA模型的基础上构建LDA-MURE模型,为用户推荐不同类别的背景音乐。实验结果表明,所提方法的MEA指标值较低、P@N指标值较高、用户满意度较高。
Different emotional states of users require different background music,so an adaptive background music recommendation method based on LDA-MURE model is proposed.The audio features and social tags of background music are extracted,the features of the above data are fused by Fisher linear discriminant analysis method,and the intra-class dispersion and inter-class dispersion of different types of background music are obtained by combining with the projection transformation method.The rhythm cycle changes of human emotions are analyzed through modern psychology,based on which the user’s current emotional state is judged.Finally,the LDA-MURE model is constructed based on the LDA model to recommend different types of background music for users.The experiment results show that the proposed method has low MEA index value,high P@N index value and high user satisfaction.
作者
杨静
YANG Jing(Shangluo University,Shangluo 726000,Shaanxi Province,China)
出处
《信息技术》
2024年第6期136-140,146,共6页
Information Technology
基金
陕西省自然科学基金(2021JM3041)。
作者简介
杨静(1979-),女,本科,讲师,研究方向为音乐教育。
