Multi-class time series classification of EEG signals with Recurrent Neural Networks

Electroencephalogram (EEG) is one of the electrophysiological tests commonly used to record electro-chemical reactions in neural network. In this process various electrodes are connected in 10-20 pattern in different points in brain, the acquisition of brain activities takes place with 16 channels or 32 channels, etc., Each channel records the information of electrode connection region as one dimensional (1D) signals. It is very important to interpret this 1D signals and classify different activities of brain for various diagnostic purpose. In this paper, different Deep learning algorithms for multiclass, time series classification of different electrical activities in brain are carried out. A comparative study between simple Recurrent Neural Network (simple RNNs), Long-Short Term Memory (LSTM) and Gated recurrent Units(GRUs) is tried out for EEG signals acquired from people having different pathological and physiological brain states.