Abstract

This paper presents affective features designed for music and develops a method to predict dynamic emotion ratings along the arousal and valence dimensions. We learn a model to predict continuous time emotion ratings based on combination of global and local features. This allows us to exploit information from both the scales to make a more robust prediction.

Date

2014

Authors

Naveen Kumar, Rahul Gupta, Tanaya Guha, Colin Vaz, Maarten Van Segbroeck, Jangwon Kim, Shrikanth S Narayanan

Conference

MediaEval