A sleep staging system and method based on a dual-stream parallel neural network
By employing a dual-stream parallel neural network architecture and feature fusion technology, the problem of difficult N1 stage identification in single-channel EEG is solved, achieving efficient and accurate sleep staging, which is suitable for health monitoring and medical auxiliary diagnostic systems.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- SHENZHEN KUKAI BRAIN MACHINE INTELLIGENT TECHNOLOGY CO LTD
- Filing Date
- 2026-03-14
- Publication Date
- 2026-06-23
AI Technical Summary
Existing automatic sleep staging technologies based on single-channel EEG have difficulty accurately identifying N1 stage due to data imbalance, low signal-to-noise ratio, and the lack of independent verification of instantaneous features in the model architecture, resulting in low recall rate for N1 stage.
A dual-stream parallel neural network architecture is adopted, including a shared feature extraction module, a context processing stream, and a local retention stream. Information is integrated through a feature fusion module, and the model training is optimized by combining a weighted cross-entropy loss function to improve the N1-period recognition capability.
It significantly improves the recognition accuracy of N1 stage while maintaining the recognition accuracy of other sleep stages, thereby improving the overall sleep staging performance. It also has high computational efficiency and is suitable for practical applications.
Smart Images

Figure CN122251027A_ABST