基于深度置信网络的音乐配乐识别研究
Research on music score recognition based on deep confidence network
焦健 1谢展鸿2
作者信息
- 1. 兰州财经大学艺术学院,兰州 730020
- 2. 梧州学院教师教育学院,广西梧州 543002
- 折叠
摘要
如何有效地管理与识别音乐元素信息是近年来数字音乐研究的重点.因此,基于深度置信网络提出一种音乐配乐识别算法,从多个角度对音乐特征信息进行提取,从音乐情感角度分析,对深度置信模型进行优化,最后将提取音乐特征数据输入到优化后的深度置信模型完成音乐的识别分类.经实验分析,在不同音乐配乐标签的测试中,相较于其他识别模型,优化后深度置信识别模型能够更准确地识别各个情感标签,均值为91.14％,识别性能表现较好,可满足音乐配乐的识别与管理要求,该研究内容可以为音乐数据的检索与管理提供技术参考.
Abstract
How to effectively manage and identify music element information is the focus of digital music re-search in recent years.Therefore,based on the deep confidence network,a music score recognition algo-rithm is proposed to extract music feature information from multiple perspectives,analyze from the perspec-tive of music emotion,optimize the deep confidence model,and finally input the extracted music feature da-ta into the optimized deep confidence model to complete the recognition and classification of music.Through experimental analysis,in the test of different music score tags,compared with other recognition models,the optimized deep confidence recognition model can more accurately identify each emotion tag,with an average of 91.14％.The recognition performance is the better,meeting the recognition and management require-ments of music score.The research content can provide technical reference for the retrieval and management of music data.
关键词
深度置信/音乐识别/情感分类/多特征/配乐标签Key words
deep confidence/music recognition/emotion classification/multiple features/music label引用本文复制引用
出版年
2024
NETL
NSTL
万方数据