A dual-stream deep neural network integrated with adaptive boosting for sleep staging

Yongkangjian Fang, Yi Xia, Peng Chen, Jun Zhang, Yongliang Zhang

Abstract: Highlights•A novel deep learning model with a dual-stream structure is proposed for sleep staging.•The model accepts 1D EEG raw signal and its 2D CWT representation in each stream.•The low and high frequency features can be extracted in the 1D signal stream.•A modified ResNet50 network with attention mechanism is utilized in the 2D signal stream.•An adaptive boosting algorithm is utilized to alleviate the class imbalance problem.