Abstracts Submitted: 510
Number of Users: 710
View Abstracts Submitted
Back to home Page
Identifying the mood or the emotion in a song is an interesting ongoing research involving different parameters of an audio sample. This paper is an attempt to gain clarity on the subject using Indian Classical music and its concepts. A Raag, which is a central concept in Indian Classical music, is said by the learned to adhere to a specific set of emotions. It is therefore hypothesized that the emotional content of an audio can be correlated to the melodic content. Thus, the moods of pop songs based on certain Raags are tested and correlated to the moods of the respective Raags. This comparison creates a requirement of identifying Raags in pop compositions. To achieve this, Machine Learning algorithms involving Support Vector Machines and Neural Networks are used to match the Pitch Class Distributions, Bi-gram Distributions, and Tri-gram Distributions of pop compositions to certain Raags present in the data set in order to identify those Raags in the compositions. The comparison also creates a requirement to test the emotional content of Raags and pop songs. It is achieved by performing self-report tests using the Circumplex Model of Affect. The Pitch Class Distribution algorithm using Neural Networks achieves 100% accuracy in identifying Raags in pop songs. The testing of moods in songs achieves 80% accuracy when comparing the Valence axis of the Circumplex Model of Affect for Raags and songs based on it.
© Copyright 2017 All Rights Reserved