Contactless Respiration Disorder and Snores Monitoring Using Advanced Signal Processing Exploiting Radar Signals

Obstructive sleep apnea (OSA) is a widespread sleep disorder affecting a minimum 4% and up to 30% of adults globally with significant health risks, such as cardiovascular diseases and cognitive impairment. Current diagnostic methods, such as polysomnography, load patients with multiple sensors, require special laboratories, and are uncomfortable for long-term monitoring. In this paper, we present a radar-based contactless system to monitor respiratory disorder patterns to detect fatal sleep problems. Using a 24 GHz continuous-wave radar, we extracted respiratory displacement waveforms and respiration rates, validated them against a ground-truth and achieved high accuracy. Following this, we applied signal processing techniques to extract key respiratory events such as apnea, hypopnea, and snores. These features were used to train a multiclass k-nearest neighbours (KNN) classification model. The model demonstrated excellent performance for the detection of apnea, hypopnea, and snores. We achieved a validation accuracy of 99.79% for respiratory displacement waveforms and 99.99% for respiration rates against the ground truth. Multiclass classification model achieved over 99% accuracy in detecting apnea, hypopnea, and snore events. The results indicate that radar-based system can be used effectively in clinical and home settings to monitor respiratory health and detect abnormal events in real time.