Fast prototyping Body Sensor Network for sleep monitoring at home
Konstanz, September 2015. There is a rich diversity of monitoring systems available on the market nowadays that promise to offer information about sleep quality of the user. Whether they are fitness trackers, smart watches, smart shirts, smartphone applications or patches, these devices do not provide access to the raw data, algorithm description or a disclosed agreement ratio with the gold standard, PSG.
Presentation of the Master Thesis
Prof. Dr. Wache, Oana-Ramona Velicu, Prof. Dr. Seepold, Prof. Dr. Martínez Madrid (l.t.r)
Also, many such systems only distinguish between sleep and wake states, or between wake, light sleep and deep sleep, and it is not always clear how these are mapped to the five known sleep stages: wake, REM, NREM1, NREM2, NREM3-4. This work presents an overview of sleep phenomenology, sleep disorders, sleep monitoring modalities and commercial sleep tracking solutions. Experiments and sleep classification algorithms published in sleep literature were revised in an attempt to find a reduced-complexity method of processing a minimum number of bio vital signals, while providing accurate results compared to the gold standard, Polysomnography (PSG). The Maser Thesis presented by Oana-Ramona Velicu shows how a body sensor network can be built as an Internet of Things object and used to monitor bio vital signals like wrist activity and heart rate. The recorded data is displayed on a web page, providing the user or a possible observer (medical staff members) with real time feedback. The thresholds used for sleep classification are dynamically adapted during every recording session, therefore the hypnogram is displayed when the session is stopped. This method was preferred, to cope with differences in vital signals characteristics due to factors like age, user weight or possibly others. This method was coupled with Kushida’s equation to classify the whole range of sleep stages, but at the moment, agreement ratio with PSG was not determined.