基于深度学习的音乐情感识别  

Music Emotion Recognition Based on Deep Learning

作　　者：唐霞 张晨曦[1] 李江峰[1] TANG Xia;ZHANG Chen-xi;LI Jiang-feng(School of Software Engineering, Tongji University, Shanghai 201804, China)

机构地区：[1]同济大学软件学院,上海201804

出　　处：《电脑知识与技术》2019年第4Z期232-237,共6页Computer Knowledge and Technology

基　　金：国家自然科学基金"基于势能导向的互联网视频资源分布式情感搜索模型研究"(项目编号:61702372)

摘　　要：随着互联网多媒体技术的发展,越来越多的音乐歌曲通过网络发布并存储在大型数字音乐数据库中。针对传统音乐情感识别模型音乐情感识别率低的问题,本文提出一种基于深度学习的音乐情感识别模型。该模型使用音乐信号特征语谱图作为音乐特征输入,使用卷积神经网络和循环神经网络相结合的方法对语谱图进行特征提取和情感分类。实验表明,相比于单独使用CNN、RNN等情感识别模型,该模型对音乐情感识别率更高,对音乐情感识别的研究具有重大意义。With the development of Internet multimedia technology, more and more music songs are issued through the Internet and stored in large digital music databases. Aiming at the problem that the emotion recognition accuracy of traditional music emotion recognition model is low, this paper proposes a music emotion recognition model based on deep learning. The model uses the music signal feature spectrogram as the music feature input, and uses the combination of convolutional neural network and recurrent neural network to extract features of spectrograms and classify their emotions finally. Experiments show that compared with the emotion recognition models such as CNN and RNN, the model has higher emotion recognition accuracy for music, which is of great significance for the study of music emotion recognition.

关 键 词：音乐情感识别 深度学习 语谱图 卷积神经网络 循环神经网络 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]