Name
Abstract | ||
|---|---|---|
The goal of music thumbnailing is to find a short, continuous segment of a song that represents the whole song. In light of the observation that a main function of music is to communicate emotions, in this short paper we investigate whether a representative part selected by some automatic mechanism for emotion recognition corresponds to the chorus section of music. To address this research question, we introduce a so-called attention layer with long-short term memory cells to a deep convolutional neural network for music emotion classification. The attention layer estimates the importance of each 3-second chunk of a song in predicting the song-level emotion labels of the song. To this end, a collection of 31K songs with emotion labels are used. We then generate a thumbnail for each song based on the importance scores and examine whether the thumbnail corresponds to any chorus section of the song, using another dataset with annotations of chorus sections. Our experiment shows that for 35% of the songs our thumbnails contain a whole chorus section, and that for 80% of the songs the thumbnails overlap 50% in time with a chorus section. |
Year | Venue | Field |
|---|---|---|
2017 | Asia-Pacific Signal and Information Processing Association Annual Summit and Conference | Thumbnail,Research question,Emotion recognition,Convolutional neural network,Computer science,Support vector machine,Emotion classification,Speech recognition,Chorus,Artificial neural network |
DocType | ISSN | Citations |
Conference | 2309-9402 | 3 |
PageRank | References | Authors |
0.38 | 6 | 3 |
Name | Order | Citations | PageRank |
|---|---|---|---|
| Yu-Siang Huang | 1 | 11 | 2.56 |
| Szu-Yu Chou | 2 | 49 | 6.82 |
| Yi-Hsuan Yang | 3 | 1022 | 84.71 |