A Hierarchical LSTM Model for Modeling EEG Non-Stationarity

Md Musaddaqul Hasib, Tapsya Nayak1, Yufei Huang1

  • 1University of Texas at San Antonio

Details

18:15 - 20:15 | Mon 5 Mar | Caribbean ABC | MoPO.14

Session: Poster Session # 1 and BSN Innovative Health Technology Demonstrations

Abstract

Recent progress in using Long Short-Term Memory (LSTM) Network in sequence-to-sequence learning of video, text, image has motivated us to explore its use Electroencephalogram (EEG) sequence signals. However, modeling an EEG sequence is a challenging task due to its high dimensionality. The goal of this current work is to predict the human decision from continuous EEG signals. An application was designed to guard a restricted area, a decision of allow or deny is made based on the physical appearance and identification card. In this paper, we propose a hierarchical long short-term memory (H-LSTM) model where the first layer encodes special correlations between each EEG channel epoch and the second layer encode temporal correlations between each epoch in a sequence. Thus, this novel approach can address non-stationarities in EEG data. The proposed model highlights the time points contributing to classification of human decision made at epoch level. Classification results of guard's decision (Allow/Deny) is reported from 18 participants. Our results indicate H-LSTM model outperforms a LSTM model by 12.4% and a shallow Support Vector Machine mode by 17.4%. Our results suggest that the H-LSTM model can be utilized effectively to predict human decision or other similar applications.