Predicting Perceived Music Emotions with Respect to Instrument Combinations

Music Emotion Recognition has attracted a lot of academic research work in recent years because it wide range applications, including song recommendation and music visualization. As is way for humans to express emotion, there need machine automatically infer the perceived emotion pieces music. In this paper, we compare accuracy difference between recognition models given as whole versus separated by instruments. To models' predictions, which are distributions over valence arousal values, provide metric that compares two distribution curves. Using metric, empirical evidence training Random Forest Convolution Recurrent Neural Network with mixed instrumental data conveys better understanding than same into each source.