Sistema automático de detecção do sono utilizando células de carga

Abstract

The quality of a person's sleep is directly associated with their health. In view of this, sleep monitoring is an excellent resource for the diagnosis and follow-up of various diseases. Polysomnography, the gold standard for sleep monitoring, is widely used in clinics specializing in sleep medicine. However, because it is an intrusive and expensive method, its continuous and long-term use becomes impracticable. Currently, there are several non-intrusive sleep monitoring techniques under study and development. They seek to remove the need for user intervention on the device. One such technique makes use of load cells positioned under the bed supports as an alternative and non-intrusive method for monitoring sleep. Information about the respiratory rhythm, heart rhythm and movements of the person can be extracted from the acquired signs, which can help in the process of detecting sleep. In view of the above, the present work presents the development of a sleep detection system based on machine learning, able to determine if an individual is in the waking state or sleep, from measurements from load cells coupled to the brackets of the bed. The system was evaluated based on the results obtained for 39 patients, resulting in a sensitivity (awake class) of 50.1% and specificity (sleeping class) of 53.1%. After applying a medium filter on the signal coming from the classification, there was an increase of 2.8% in the sensitivity that remained in 51.5%, while in the specificity was obtained 54.8%, which represents an increase of 3.2 % (sic).