基于支持向量机（SVM）的音乐流派分类
摘要随着数字音乐产业的发展和网络技术的提升，人们可以通过各种音乐网站访问容量巨大的数据库。为使用户能更容易寻找到自己想要的音乐资源，对音乐资源进行分类，成了国内外许多学者研究的热点问题。43213
本文使用支持向量机为分类器，进行音乐分类。音乐分类的重点在于提取的特征向量能有效的表现不同音乐的特点与差异。根据短时分析技术，对音乐预处理，然后提取音乐的时域与频域特征，组成高维特征向量。使用不同组成的高维音乐特征向量和不同数据集对分类性能进行验证。实验结果显示使用MFCC参数、短时过零率和短时能量组成的38维特征向量，在GTZAN音乐库进行十流派分类，分类正确率为75.5%。在自己建立的四种中国传统乐器音乐库的分类正确率为94.375%。
关键词 音乐分类 特征提取 支持向量机 MFCC
毕业论文设计说明书外文摘要
Title Music genre classification based on support vector machine (SVM)
Abstract
With the enhancement of the digital music industry and the development of network technology, people can access a huge database capacity through various music sites. To make it easier for users to find the resources they want music, the classification of music resources is becoming a hot issue for many scholars at home and abroad.
In this paper, the support vector machine is used as a classifier to classify music.Music classification is focused on the extracted feature vector can effectively show the characteristics and difference of different music . Use music pretreatment, according to the short-term analysis technology, and then extract the time domain and frequency domain characteristics of music, form a high-dimensional feature vector. Using a higher dimensional music of different characteristic vector and the different data sets to verify this classification performance. Experimental results show that using MFCC parameters, of short-time energy and short-time zero crossing ratio 38 dimensional feature vector, the GTZAN dataset ten genre classification, classification accuracy of 75.5%. Four kinds of Chinese traditional music instruments in their library classification accuracy of 94.375%.
Keywords Music classification,Feature Extraction,SVM,MFCC
目 录
1 绪论 1
1.1 研究的背景和意义 1
1.2 音乐流派分类的现状 1
1.3 本文主要研究内容和结构安排 3
2 音乐理论基础 4
2.1 人耳的感知 4
2.1.1人耳的构造 4
2.1.2 听觉感受性 4
2.1.3 掩蔽效应 6
2.2 音乐的特征 7
3 支持向量机基本理论 8
3.1 统计学习理论 9
3.2 支持向量机分类 10
3.2.1 最大间隔分类器和硬间隔分类法 10
3.2.2 线性不可分支持向量机和软间隔分类法 11
3.2.3 非线性支持向量机与核函数 12
3.3 支持向量机的特点和优势 13
4 音乐特征的提取 15
4.1 信号预处理