A novel approach for detection of medial temporal discharges using blind source separation incorporating dictionary look up

In blind source separation (BSS), sparsity is proved to be very advantageous. If data is not sparse in its current domain, it can be modelled as sparse linear combinations of elements of a chosen dictionary. The choice of dictionary that sparsifies the data is very important. In this paper the dictionary is pre-specified based on chirplet modelling of various kinds of real epileptic spikes. Dictionary look up together with source separation is used to extract the closest source to the source of interest from the scalp EEG measurements. The algorithm has been tested on synthetic and real data consisting of epileptic discharges, and the results are compared with those of traditional BSS.