A Song Clustering Method Based on Iterative K-means Algorithm

Abstract

The invention discloses a song clustering method based on iterative k-means algorithm, extracts MFCC vectors of songs in a music library as acoustic features, and uses iterative k-means algorithm to perform emotion recognition and emotion classification on songs, specifically including the following Steps: extract the MFCC of the songs in the music library, and obtain the MFCC vector of each song; label all the MFCC vectors with song IDs and collect them into a data set; perform the first k-means clustering on the data set to obtain each MFCC vector Clustering results; generate a K-dimensional label vector set according to the proportion of the MFCC vectors contained in the songs belonging to each cluster; perform the second k-means clustering on all K-dimensional label vector sets to obtain the final song clustering result. The present invention can classify the massive songs in the music library according to the similarity of the emotions perceived by human ears among the songs, so as to recommend similar songs to users according to the emotion classification more accurately and effectively.

Description

technical field [0001] The invention relates to the technical field of data processing, in particular to a song clustering method based on an iterative k-means algorithm. Background technique [0002] In today's Internet era, large music portals have song libraries with massive data. Users often have the need to find songs that are similar to or belong to the same category as their favorite songs. Traditional search engines are only suitable for searching under the condition that the user has a clear understanding of the target song information, and cannot search based on the similarity of the emotions perceived by the human ear between songs. The traditional method of classifying music by manually setting music tags is inefficient, not suitable for processing massive data, and the accuracy of manual tags is low. Therefore, the music portal websites in the existing environment lack efficient and accurate methods for classifying massive songs, resulting in users being unabl...

AI Technical Summary

Problems solved by technology

Traditional search engines are only suitable for searching under the condition that the user has a clear understanding of the target song information, and cannot search based on the similarity of the emotions perceived by the human ear between songs

The traditional method of classifying music by manually setting music tags is inefficient, not suitable for processing massive data, and the accuracy of manual tags is low

Therefore, the music portals in the existing environment lack efficient and accurate methods for classifying massive songs, which makes it impossible for users to quickly and easily find songs that they may be interested in based on the similarity between the rest of the songs and their favorite songs.

Images