SalientSleepNet

Code for the model in the paper SalientSleepNet: Multimodal Salient Wave Detection Network for Sleep Staging (Accepted by IJCAI 2021).

Datasets

We evaluate the performance of SalientSleepNet on Sleep-EDF-39 and Sleep-EDF-153 datasets, which are open-access databases of polysomnography (PSG) recordings.

Requirements

• Python 3.7
• Tensorflow 2.3.1

Function of file

• construct_model.py
  • Generate submodule of model (E.g., U-unit, MSE).
• model.py
  • Generate single-branch and two-branch model.

Usage

We've offered two models: SingleSalientModel and TwoSteamSalientModel for single-branch method and two-branch method respectively. You need to use a tensor with shape: [None, group_sequence_number , group_sequence_length * sleep_epoch_length, 1, 1] for the single-branch method while duplicate tensors with the same shape for the two-branch method.