1. Adaptive deep knowledge and self supervised framework for classifying sleep stage using neural network.
- Author
-
Victor, Ceronmani Sharmila, Madanuru, Manideep Sai, Goodalvadi, Jagadeesh, and Samanu, Sai Prathap
- Subjects
SLEEP stages ,DEEP learning ,TRANSFORMER models ,THEORY of self-knowledge ,PUBLIC health ,SLEEP deprivation - Abstract
Sleep disorders are a significant public health issue for humans. To monitor these disorders, for autonomic sleep staging, an EEG signal is currently thought of as standard. The clinical setup for EEG sleep stage is very expensive. Also, extreme care should be taken while setting up the EEG. Hence electrocardiogram (ECG) is chosen over EEG results in increment of performance of ECG. Deep learning is EEG based sleep stage classification has found best results recently. The process of these models involves training, testing the huge amounts of data which can be produced from sleep labs but labelling these data is a difficult task. The proposed model results in good competitive performance besides eliminating the black-box behaviour of DL models with the help of interpretability aspect of the attention modules. The designed cross-modal transformers are made up of all original cross-modal neural network design and a multistage 1-D neural network for automatic representation learning. Our design results in better sleep stage classification performance with interpretability, along with a fourfold reduction in the number of parameters. [ABSTRACT FROM AUTHOR]
- Published
- 2024
- Full Text
- View/download PDF