| Insomnia or low sleep quality has become a major problem affecting the health of modern people. Accurate sleep staging is a prerequisite for a correct understanding of sleep architecture, sleep quality, and relevant mechanism. Most of the traditional methods of automatic sleep staging make use of electroencephalogram(EEG) signals, however, the acquisition of EEG signals requires sophisticated instruments and complex electrodes, which cannot meet the needs of sleep monitoring at home. Pressure sensing electric mattress, which minimize the impact of sleep when sleep monitoring, is very suitable for sleep monitoring at home. But BCG signal acquired by the mattress is mixing signal of heartbeat, respiration and body movement. Body movement signals, which produce effects on the extraction of heartbeat and respiration signals characteristics, cannot be ignored. Furthermore, the body moving signal itself also has significant value to sleep staging. Most of the non EEG automatic sleep staging algorithm is based on a single kind of signal and many methods employ standard physiological signals collected by polysomnography. These methods may not perform high accuracy using prolonged pressure electric hybrid signal. This paper presents a multiple physiological signals sleep staging algorithms, classifying sleep stage based on mattress acquisition of piezoelectric data. Meanwhile, we also realize a mobile sleep monitoring system based on piezoelectric signals acquired by mattress and this system provides a scientific solution for home sleep monitoring and remote sleep monitoring.The multiple physiological signals sleep staging algorithm, for the first time, applies hidden Markov model to the respiratory signal of the sleep staging. The attempt demonstrates hidden Markov model can identify different modes of breathing in different sleep phases. By analysis of heart rate and respiratory rate of hidden Markov model, we find the advantages and disadvantages of the two models when identifying different sleep stages and combines the advantages to improve the original algorithm. Motion detection algorithm is also designed to assist the algorithm, which corrects the wrong staging affected by movement activity. The algorithms can automatically identify the awakening period, light sleep, deep sleep and REM periods and body movement number is also recorded. The result achieves a total accuracy of 66.91, and the algorithm have better performance in the comparison of similar products.In this paper, a software platform of mobile sleep monitor based on piezoelectric sensing is set up, which provides a scientific solution for the realization of home sleep monitoring and remote sleep monitoring. The software based on the Android platform and web technology defined sleep physiological data communication scheme and the user’s interaction, integrating improved Markov model of sleep staging algorithm and preprocessing of BCG signals. The system realized the functions of remote sleep monitoring, visualization of sleep structure, sleep quality analysis, social network for ward mates, and knowledge model for people with insomnia. |
