COUPLED DICTIONARY LEARNING FOR MULTIMODAL DATA: AN APPLICATION TO CONCURRENT INTRACRANIAL AND SCALP EEG

摘要

This paper focuses on learning a coupled dictionary between multimodal datasets where the data of different modes can be described as a function of each other. Our method is able to reconstruct the data of one mode by using the data of another mode. This provides the advantage on applications that low-quality data are generally available and high-quality data are not. We employ a concurrent intracranial and scalp EEG dataset, to learn a dictionary and a mapping function between the two modalities. The aim is to infer the intracranial from only the scalp EEG by using that dictionary and mapping function. The novelty of this work is the development of an algorithm that obtains an optimal coupled dictionary, sparse coefficients and the mapping function between modalities.