Sleep monitoring algorithm based on multi-modal sensor fusion

Abstract

Polysomnography (PSG) is the gold-standard for sleep monitoring, but because of the obtrusiveness of its attached sensors, non-invasive sleep monitoring algorithms have been developed throughout the years. Nonetheless, the previous research are not proven to be reliable whereas, most of the products are designed based on controlled healthy subjects. We present a novel sleep stages classification algorithm using low cost and noncontact multi-modal sensor fusion by: extracting vital signals from radar signals; and sound-based context-awareness technique. This work is designed based on the PSG database of sleep disorder patients, received and certified by sleep technician and medical doctor at Hanyang University Hospital. The proposed algorithm further merges medical/statistical knowledge to personalize the algorithm, and devise post-processing. The efficiency of the proposed algorithm is viewed by contrasting sleep scoring performance between single sensor and sensor-fusion algorithms. To verify the possibility of commercializing this algorithm, the classification results of this algorithm were further contrasted with the commercialized sleep monitoring device, ResMed S+. The proposed algorithm was experimented with random patients who took PSG exams, and results show a encouraging approach for monitoring sleep stages.