Comparison of Functional Connectivity and Multivoxel Pattern Analysis for Classification of Musical Genres

Abstract

Humans often tend to group auditory stimuli into discrete categories, including categories such as animal species, language, musical instruments, and musical genres. Among these classifications, music genre emerges as a common dimension in assessing human music preferences. Neuroimaging studies have reported a correlation between brain activities and musical genres but there is a little study to investigate performance of functional connectivity and multivoxel pattern analysis for the classification of the musical genres.

In this work, we used fMRI data of 5 subject in 6 sessions during listening to different musical genres to find a relation between music genres and functional connectivity and multivoxel pattern analysis (MVPA). To this end, we extracted features from the audio files and explored their relation with brain activity. We obtained promising  results for MVPA with 74.3% accuracy but no significant or separable result for functional connectivity.  Voxel averaging during the parcellation process may have removed some important information useful for the classification of the musical genres.