Abstract

The emotional content perceived from music has great impact on human beings. Research related to music is attaining more and more recognition not only in the field of musicology and psychology but also getting attention of engineers and doctors. The categorization of music can be carried out by considering various attributes such as genres, emotional content, mood, instrumental etc. In this work Hindi music signals belonging to different genres are categorized in four quadrants belonging to Arousal-Valence (AV) plane of emotion categorization i.e. Positive Valence-Positive Arousal (PV-PA), Positive Valence- Negative Arousal (PV-NA), Negative Valence-Positive Arousal (NV-PA) and Negative Valence-Negative Arousal (NV-NA). Features related to music are calculated by using MIR toolbox and the classification techniques used are K-Nearest Neighbor (K-NN), Naive Bayes (NB) and Support Vector Machine (SVM). In this work authors make use of two types of annotation techniques i.e. Subject Based Annotation (SBA) and Algorithm Based Annotation (ABA). The accuracy, precision and recall are considered as evaluation parameter in this work. The evaluation parameters for all the classification techniques using both the annotation methods are compared in the proposed work. Results reveal that SVM classifier outperforms other two classifiers in terms of the parameters considered for both SBA and ABA and it has also been proved that algorithmic annotation outperforms subjective annotation.