Multi Channel Brain FEG Signals Based Emotional Arousal Classification with Unsupervised Feature Learning using Autoencoders

The importance of learning important features in an automatic manner is growing exponentially, as the volume of data and number of systems using pattern recognition techniques continue to increase. In this paper, arousal recognition from multi channels EEC signals was conducted using human crafted statistical features and learned features from 32 different EEG source channels. We have obtained 98.99% accuracy rate with unsupervised feature learning approach for Arousal classification. Unsupervised feature learning worked better compared to handcrafted feature approach.