期刊
IEEE SIGNAL PROCESSING LETTERS
卷 24, 期 8, 页码 1208-1212出版社
IEEE-INST ELECTRICAL ELECTRONICS ENGINEERS INC
DOI: 10.1109/LSP.2017.2713830
关键词
Convolutional neural networks; feature aggregation; music auto-tagging; transfer learning
资金
- Korea Advanced Institute of Science and Technology [G04140049]
- National Research Foundation of Korea [2015R1C1A1A02036962]
- Ministry of Science, ICT & Future Planning, Republic of Korea [G04140049] Funding Source: Korea Institute of Science & Technology Information (KISTI), National Science & Technology Information Service (NTIS)
- National Research Foundation of Korea [31Z20130012985] Funding Source: Korea Institute of Science & Technology Information (KISTI), National Science & Technology Information Service (NTIS)
Music auto-tagging is often handled in a similar manner to image classification by regarding the two-dimensional audio spectrogram as image data. However, music auto-tagging is distinguished from image classification in that the tags are highly diverse and have different levels of abstraction. Considering this issue, we propose a convolutional neural networks (CNN)-based architecture that embraces multi-level and multi-scaled features. The architecture is trained in three steps. First, we conduct supervised feature learning to capture local audio features using a set of CNNs with different input sizes. Second, we extract audio features from each layer of the pretrained convolutional networks separately and aggregate them altogether giving a long audio clip. Finally, we put them into fully connected networks and make final predictions of the tags. Our experiments show that using the combination of multi-level and multi-scale features is highly effective in music auto-tagging and the proposed method outperforms the previous state-of-the-art methods on the MagnaTagATune dataset and the Million Song Dataset. We further show that the proposed architecture is useful in transfer learning.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据