Techniques and Algorithms for the Characterization of Sleep Quality through Wearable Devices

This thesis work was developed with the purpose of creating an innovative home sleep monitoring system using two devices designed by STMicroelectronics: the BodyGateWay patch, and the neMEMSi inertial platform. The BGW allows to detect ECG, accelerometer and bioimpedance signals while neMEMSi returns an estimation of the orientation of the platform. The devices are non-invasive, wearable and with Bluetooth communication capabilities. The first step was to determine the best device placement in order to extrapolate the most useful range of information regarding the quality of the patients sleep. The setup chosen enables to detect events of sleep apnea, periodic limb movement, and the sleep staging. Three algorithms have been developed. The first one is able to determine, from the accelerometer signal, if the subject is sleeping or not. The second performs, with a specifically designed Hidden Markov Model, the various sleep stages. The third algorithm processes the Electroculogram signal to identify the REM sleep stages - it is used to check if the second algorithm works properly. The promising outcome of this work suggests the potential use of this system as a support for physicians who must decide whether to refer the patient to a full polysomnographic study or not