Hybrid spiking neural network for sleep electroencephalogram signals

Ziyu Jia, Junyu Ji, Xinliang Zhou, Yuhan Zhou

Abstract: Sleep staging is important for assessing sleep quality. So far, many scholars have tried to achieve automatic sleep staging by using neural networks. However, most researchers only perform sleep staging based on artificial neural networks and their variant models, which can not fully mine and model the bio-electrical signals. In this paper, we propose a new hybrid spiking neural network (HSNN) model for automatic sleep staging. Specifically, we use a spiking neural network to classify sleep EEG signals. In addition, we adopt a hybrid macro/micro back propagation algorithm, aiming to overcome the limitations of existing error back propagation methods for spiking neural network. In order to verify the effectiveness of HSNN, we evaluate it on the public sleep dataset ISRUC-SLEEP (Institute of Systems and Robotics, University of Coimbra-Sleep). The results show that the proposed method achieves satisfactory performance on ISRUC-SLEEP.