Extracting Vital Sings from Seismic Signals
Background:
Sleep monitoring is a crucial aspect of health tracking, significantly aiding in assessing a person’s overall well-being. Traditional monitorization methods utilize respiratory tracking through uncomfortable breathing devices and heartbeat rate (HR) measurement via body-contact wearables like chest straps and wrist sensors. These methods often cause discomfort or are neglected at bedtime due to their intrusive nature, particularly among older adults or patients. Innovatively, biomedical vibration signals, such as seismocardiogram (SCG) and ballistocardiogram (BCG), which measure micro-vibrations produced by the heart movement, are analysed for human health assessment and monitoring. SCG, capturing the micro-vibrations from heart movements, has been used to implement user authentication on mobile phones by analysing heartbeat-induced chest vibrations. Understanding and analysing vital parameters like heart rate and respiration are essential for monitoring cardiac health and diagnosing heart-related diseases. As a result, a variety of wearable healthcare devices have been developed for continuous electrocardiogram (ECG) monitoring, catering to both patients and health-conscious individuals. Despite their benefits, the need for constant body attachment makes these devices cumbersome, particularly during sleep, highlighting the demand for less obtrusive, more user-friendly monitoring technologies. There is, therefore, a need for non-intrusive solutions that continuously monitor cardiac events during night periods.
Aim:
Development of algorithm(s) to extract sleep-related activities such as heart rate, respiration rate, and activity pattern (e.g., leg movements) in bed from seismic signals of individuals lying on a bed. Additionally, if the student shows interest, the algorithms may be extended to estimate sleep positions and basic sleep parameters.
Material and Methods:
The student will begin with a literature review on algorithmic methods to derive HR, RR, and motion from seismic signals. Building on this, they will implement and test algorithms capable of extracting the parameters of interest. Data will be provided and collected from an ongoing study, where seismic signals are recorded simultaneously with reference systems (e.g., ECG and commercial sleep monitoring devices). This will allow evaluation of algorithm performance by comparing extracted parameters against ground truth. Depending on interest, additional algorithm development may explore multi-sensor fusion, advanced signal decomposition, or machine learning–based classification approaches for sleep staging or posture recognition.
Nature of the Thesis:
Development of the algorithms: 60%
Data analysis: 30%
Data collection: 10%
Requirements:
- Good programming skills
- Good knowledge of signal processing
- Basic knowledge of data analysis
Supervisors:
Prof. Dr. Tobias Nef
Noora Angelva, MSc
Institute:
ARTORG Center for Biomedical Engineering Research, University of Bern, Gerontechnology and Rehabilitation Group