Music Emotions Recognition Based on Feature Analysis

Abstract

Music emotions recognition (MER) is a challenging field of studies addressed in multiple disciplines such as musicology, cognitive science, psychology, arts and affective computing. In this paper, music emotions are classified into four types known as exciting, happy, serene and sad. MER is formulated as a classification problem in cognitive computing where music features are extracted. And, the feature sets are input into Support Vector Machine (SVM) and Convolutional Neural Networks to classify the music emotion. It can be seen that the best accuracy of 88.2% in VGG16 where Chirplet has been turned into features images. The results show that the feature graph is feasible for music emotion classification.