User Customization for Music Emotion Classification using Online Sequential Extreme Learning Machine

Abstract

Music is an art composed by sound. Music emotion recognition as a research topic stands on different areas such as psychology, musicology. The purpose of this work is to give a recommendation of music to the user by recognizing music emotion using machine learning algorithm. In order to take the music emotion recognition, a set of musical characteristics generated by MIR Tool Box has been used. Several machine learning algorithms are used and compared in this work. For traditional method such as k-nearest neighbour classifier (k-NN classifier) and state-of-the-art neural network such as support vector machine (SVM) and extreme learning machine (ELM). For the recognition result, it cannot get a full accuracy for every user. To improve the result, the online sequential extreme learning machine (OSELM) is used to learn one by one with a fixed size of new data for the user reported result then updating the model using the latest data.