DNN Based Music Emotion Recognition from Raw Audio Signal

Music emotion is an important component in the field of music information retrieval and computational musicology. Considering the complexity of the state-of-the-art methods and challenges brought by lack of information available in existing public databases, we propose novel approach to emotion recognition based on convolutional and recurrent neural networks for feature mining. The proposed method utilizes a latest findings in deep learning by stacking convolution layer with bidirectional gated recurrent unit. The method was evaluated on the MediaEval Emotion in Music dataset and has shown exceptional performance using only raw audio signals and without any need for pre-processing.