Synchronization Measures of Neuro-Statistical Based Selected EEG features

Electroencephalographic (EEG) signals are the recordings of brain?s spontaneous electrical activity along the scalp. It may be low or high dimensional data according to the numbers of electrode placed on the scalp. High dimensional EEG data take longer time to analyze. So features are searched with reduced dimension using neural canonical correlation analysis (NCCA) for minimizing computational cost. The NCCA takes the advantages of neural network with CCA where data are fed sequentially without at once. In this paper, we measure capability of NCCA network for finding salient features by using various synchronization measures, namely cross correlation, coherence function, standard deviation and interdependencies. All of these measures give a useful quantification. These measures are done for selected feature sets and original dataset, which shows almost identical result. So we may claim that NCCA is a better network for feature selection (FS) of EEG signals.