Survey on Features and Classification Techniques in Music Genre Classification

Author Name(s): Swati A. Patil, K.Thirupathi Rao, Sonal Patil
Author Email:


Music is an integral part of society and even human being. A human being listen music to make a mood and change a mood. He listens music in sad and happy moments even when he is getting bored. In early days only folk music or local music is the way for this purpose. But in digital era or due to entry of various cultures in the society people are listening music from various parts of the country or world. The choice of listening the music varies from people to people from mood to mood. Sometimes we want to listen silent music or rock or pop etc. and sometimes it depends on the occasions also. As far as mood is concern people listen sad or happy or romantic music to make the mood or change the mood. This type of classification of songs is called genres. Genre classification is one of the prominent fields in MIR. Most of Genre classification involves feature extraction and the classification techniques along with some preprocessing if needed. In this review paper we tried to discuss and compare various techniques of features extraction and classification methods. In the literature most widely used features are MFCC and timbre features and classification methods are KNN and SVM. Low level feature and high level features are taken into consideration and most suitable feature is identified.


Nowadays music or songs are widely disseminated data on the Internet and the amount of uploading or releasing of music/ songs are going just increasing and not stopping or reducing. Effective organization and process such large data as well as efficient indexing, searching and retrieval of such large amount of musical data is a big encounter. Most music classification methods are on music categorized on genre, mood, or multiple semantic tags [17]. Music Genre is an important aspect used to categorize the music/songs at music shops, broadcasting and searching. Automatic music genre classification is a fundamental component of music information retrieval systems [10].


In this paper we have survey different features of any audio signal for genre classification and various machine learning techniques to classify genres. Most popular feature of audio is MFCC since it is very close to human auditory system. SVM and KNN are most widely used classification techniques in Genre Classification. All the work is on some standard datasets which are in English songs.

445 total views, 1 views today

Download PDF File

About the author: admin