ANALYSIS ON MUSIC CONTENT CORRELATION BASED ON AUDIO CHARACTERISTICS WITH K-MEANS
Music is one of the forms of culture that have exists in the human civilizations since millenniums ago. Emergence of a new era, the growth of technologies, and human’s creativity results in the creation of many variants of music. In fact, it is so variant that people start to create a new term ca...
Saved in:
| Main Author: | |
|---|---|
| Format: | Final Project |
| Language: | Indonesia |
| Online Access: | https://digilib.itb.ac.id/gdl/view/75332 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Institut Teknologi Bandung |
| Language: | Indonesia |
| Summary: | Music is one of the forms of culture that have exists in the human civilizations since
millenniums ago. Emergence of a new era, the growth of technologies, and human’s creativity
results in the creation of many variants of music. In fact, it is so variant that people start to
create a new term called genre to classify music. However, this classification is done by human
in a subjective manner. Occasional dispute and disagreement regarding a genre of a certain
song occurs between both layman or musical experts. This issue sprouts an idea to categorize
music in an objective way instead of a subjective one. In order to do that, experiment will be
done to group a set of similar music content by their audio characteristics. Using a dataset of
music that have been sound processed beforehand, the dataset will have its audio characteristics
in a quantitative value. This audio characteristics will be used to group similar songs together
in a cluster using K-means clustering algorithm. Humans tends to determine a genre of a certain
music by the atmosphere or the vibe when they hear the music. Hypothetically, the audio
characteristics take part in building that atmosphere. Therefore, this research will attempt to
prove if grouping a set of songs into clusters depending on their audio characteristics and
whether the genre labelling made by human is accurate enough. Moreover, the clustering model
will also be used to build a recommender system that will output music with the closest distance
in the same cluster in relation to the input. |
|---|