Transformer-based Low-interference Respiratory Signal Fusion Sleep Staging Algorithm

Sleep is crucial to human health, and monitoring sleep stages plays a key role in assessing both sleep quality and overall health. Traditional sleep monitoring methods primarily rely on contact-based sensors or devices, which somewhat limits their widespread application in daily life. In recent years, low-interference sleep state detection technologies have garnered increasing attention due to their ability to overcome the limitations of traditional methods, offering new possibilities for real-time monitoring. This study proposes an automatic sleep staging algorithm based on a Transformer model, which enhances classification accuracy by effectively integrating both time-domain and frequency-domain information from respiratory signals. We utilized the ISRUC-S3 dataset, specifically using abdominal respiratory signals from healthy subjects as experimental data, and performed sleep staging according to the American Academy of Sleep Medicine (AASM) standards. The experimental results demonstrate that the proposed algorithm improves classification accuracy to 53.60%, validating its potential in the field of low-intrusion sleep monitoring. This provides significant technical support for the future development of home sleep monitoring systems and smart healthcare devices.