Building convolutional neural networks (CNNs) for Music tagging.
We would like to train CNN to perform automatic music tagging, which is a classification task that aims to predict music tags using the audio signal. Meanwhile, subjects perform the same classification task, in which neural responses (fMRI/MEG) were recordered when they were presented a piece of music clips. Then we would like to compare representations of CNN with the neural representations to examine the streams of audio processing in human brain.
A lots of work still needs to be done.............
Reference: Choi, K., Fazekas, G., & Sandler, M. (2016). Automatic tagging using deep convolutional neural networks. arXiv preprint arXiv:1606.00298.