Predicting Viewer’s Preference for Music Videos Using EEG Dataset

We study to predict a viewer’s preference for a music video through the recorded EEG signals in the DEAP dataset. The experimental results show that using full-length EEG recordings the accuracy reaches 70% with 12-channel EEG. However, if the EEG is reduced to 5 channels with ICA (independent component analysis), only a few EEG segments are needed to achieve even higher accuracy, up to around 75% with the use of an autoencoder network.