计算机音乐分类辨识研究
毕业论文关键词 分类辨识 特征提取 支持向量机
Title Research of Computer Music Classification
Abstract That the rapidly growing Internet and mass storage has brought a huge amount of multimedia. Audio applications, especially speech recognition, content-based music retrieval and so on, are all being carried out research. The importance of intelligent classification of computer music is also being noticed. Classification’s development and achievements will have a great impact on music search, music recommendation and storage management, and it does have its practical significance. To try our best to reduce labeling by human labor, we need to improve the speed and accuracy of automatic classification, which requires extracting various features that can represent the music segments and establish an appropriate machine learning algorithm to build the model. However, music has its persity and ambiguity, and the classification criteria vary. But just take classification results into account, the closer to the human auditory system, the more excellent the system is. This paper is mainly about extractions of different features, and then gives a comparative analysis on the ability of recognition of distinct features, and finally select and combine those who has better performance as the input data of SVM to build the classification system.
Keywords feature extraction music classification SVM
目 次
1 绪论 ·· 1
1.1 研究背景 1
1.2 研究现状 1
1.3 论文结构及内容概述 · 2
2 人耳对音乐的认知机理 4
2.1 听觉的感知机制 4
2.2 听觉特性 4
3 音乐分类的情感模型 · 6
3.1 Hevner 模型 ·· 6
3.2 Russell 模型 · 7
3.3 Thayer 模型 ·· 8
4 音乐文件格式 10
4.1 MIDI · 10
4.2 WAV/AU 10
5 音乐的辨识特征 ·· 13
5.1 预处理 13
5.2 特征表示 ·· 15
6 辨识分类方法 20
6.1 支持向量机 · 20
7 模型训练及辨识过程 25
7.1 SVM 分类器 · 25
7.2 样本数据集 · 25
7.3 特征的宏观意义 ·· 26