基于前向神经网络的音乐情感识别  

Recognition of Music Emotion Based on Forward Neural Network

作　　者：赵薇[1] 王立昊 黄敬雯 周义楠 ZHAO Wei;WANG Li-hao;HUANG Jing-wen;ZHOU Yi-nan(Faculty of Science and Technology,Communication University of China,Beijing 100024,China)

机构地区：[1]中国传媒大学理工学部,北京100024

出　　处：《中国传媒大学学报（自然科学版）》2018年第4期1-5,31,共6页Journal of Communication University of China：Science and Technology

基　　金：中央高校基本科研业务费专项资金(3132018XNG1858);双一流:新时代青年拔尖科研人才支持项目(CUC18QB46)

摘　　要：情感是音乐的固有属性。本文通过对基于音乐多特征的比较,最终选择旋律走势作为情感特征,介绍了前向神经网络运用于音乐情感识别的基本原理,对使用前向神经网络(Feed-forward neural network)进行音乐情感识别的结果进行分析。实验结果显示:前向神经网络在音乐情感识别领域有着较高的准确性,能够有效地对音乐情感进行识别。The most common way to retrieve and describe music is emotional words.Emotion is an inherent attribute of music.Based on the comparison of multi-features based on music,the paper finally selects the melody trend as the emotional characteristics,introduces the basic principles of the forward neural network used in recognition of music emotion,and analyzes the result that uses feed-forward neural network to identify the emotion of music.The experimental results show that the forward neural network has a high accuracy in the field of emotional emotion recognition,and can effectively identify the emotion of music.

关 键 词：情感识别 前向神经网络 特征提取 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]