Non-Contact Sleep Apnea Monitoring System Based on Multi-Point Scattering Model and Multi-Scale Fusion Network

Sleep Apnea (SA) is a common disorder that can lead to decreased blood oxygen saturation and is closely associated with the development of numerous cardiovascular and metabolic diseases. Early screening is crucial for timely intervention. However, the complexity and discomfort of traditional polysomnography (PSG) limit its applicability in large-scale screening and prognosis tracking. This study proposes a non-contact sleep apnea monitoring method based on Frequency-Modulated Continuous Wave (FMCW) radar. The method is built upon a multi-point scattering model of the human body, using FMCW radar to capture chest and abdominal respiratory movements. A multi-scale fusion convolutional neural network is then employed to achieve high-precision identification of sleep apnea events. We collected synchronized radar and PSG data from 35 subjects for system training and validation. Experimental results demonstrate that the system achieves a classification accuracy of 91.37%. This system can be deployed in sleep laboratories or home environments, providing a convenient and non-intrusive solution for large-scale screening and prognosis tracking of sleep apnea.