hrvatski

MACHINE LEARNING MODEL FOR THE CLASSIFICATION OF MUSICAL COMPOSITION GENRES

The aim of this paper is to validate the applicability of standard machine learning models for the classification of a music composition genre. Three different machine learning models were used and analyzed for best accuracy: logistic regression, neural network and SVM. For the purpose of validating each model, a prototype of a system for classification of a music composition genre is built. The end-user interface for the prototype is very simple and intuitive ; the user uses mobile phone to record 30 s of a music composition ; acquired raw bytes are delivered via REST API to each of the machine learning models in the backend ; each machine learning model classifies the music data and returns its genre. Besides validating the proposed model, this prototyped system could also be applicable as a part of a music educational system in which musical pieces that students compose and play would be classified.

Machine learning ; Musical composition genre classification ; Logistic regression ; SVM ; Neural networks

nije evidentirano