Prediction of Successful Memory Formation during Audiovisual advertising using EEG signals

The prediction of memory performance using EEG signals is an active research area in Passive Brain Computer Interfaces community. In this type of prediction problem it is important to be able to use unlabeled data and to tackle the unbalanced nature of the data. In this work we propose two new Sparse Representation Classification schemes that are able to address the above properties of the data. The proposed classifiers have been tested in an EEG dataset related to neuromarketing. In the data analysis, we define a binary classification problem using EEG signals corresponding to the condition of remembering and forgetting. Furthermore, we compare the proposed classifiers with well-known classifiers. The obtained results show all classifiers perform above chance level, and, among them the proposed classifiers present the best performance in terms of Fscore and Kappa Value.