Monitoring Sleep with Force Sensor Measurement

This thesis presents methods for comfortable sleep measurement at home. Existing medical sleep measurement systems are costly, disturb sleep quality, and are only suited for short-term measurement. As sleeping problems are affecting about 30% of the population, new approaches for everyday sleep measurement are needed. We present sleep measurement methods that are based on measuring the body with practically unnoticeable force sensors installed in the bed. The sensors pick up forces caused by heartbeats, respiration, and movements, so those physiological parameters can be measured. Based on the parameters, the quality and quantity of sleep is analyzed and presented to the user. In the first part of the thesis, we propose new signal processing algorithms for measuring heart rate and respiration during sleep. The proposed heart rate detection method enables measurement of heart rate variability from a ballistocardiogram signal, which represents the mechanical activity of the heart. A heartbeat model is adaptively inferred from the signal using a clustering algorithm, and the model is utilized in detecting heartbeat intervals in the signal. We also propose a novel method for extracting respiration rate variation from a force sensor signal. The method solves a problem present with some respiration sensors, where erroneous cyclicity arises in the signal and may cause incorrect measurement. The correct respiration cycles are found by filtering the input signal with multiple filters and selecting correct results with heuristics. The accuracy of heart rate measurement has been