基于深度学习的音乐特征提取
首发时间:2014-12-24
摘要:深度学习作为一种新的特征提取技术,在语音信号处理领域取得了一系列成功。本文借鉴深度学习在语音信号处理上的研究成果在音乐分类与深度学习理论相结合的基础上,针对如何利用深度学习强大的特征提取功能发现更加适用于音乐分类的声学特征这一问题展开研究,分别采用了深度神经网络提取的特征和传统的MFCC 对音乐片段进行歌曲风格分类,在分类正确率方面提高了20%。
For information in English, please click here
Deeplearning based Music Feature Extraction
Abstract:As a new theory of feature extraction,deep learning models have achieve great success.This thesis is based on the combination of speech recognition and deep learning theory,aims at finding more suitable features for music classification task.In our experiment,we used the MFCC feature and the new features learned by deep learning model to run the classification task,have increased the classification accuracy by 10% , compared with MFCC features.
Keywords: Pattern Recognition Music Recommendation Deep learning Feature Extraction
论文图表:
引用
No.4624251102402414****
同行评议
共计0人参与
勘误表
基于深度学习的音乐特征提取
评论
全部评论0/1000