Computational Sleep Behaviour Analysis and Application
Sleep affects a person’s health and is, therefore, assessed if health problems arise. Sleep behaviour is monitored for abnormalities in order to determine if any treatments, such as medication or behavioural changes (modifications to sleep habits), are necessary. Assessments are typically done using two methods: polysomnography over short periods and four-week retrospective questionnaires. These standard methods, however, cannot measure current sleep status continuously and unsupervised over long periods of time in the same way home-based sleep behaviour assessment can. In this work, we investigate the ability of sleep behaviour assessment using IoT devices in a natural home environment, which potential has not been investigated fully, to enable early abnormality detection and facilitate self-management. We developed a framework that incorporates different facets and perspectives to introduce focus and support in sleep behaviour assessment. The framework considers users’ needs, various available technologies, and factors that influence sleep behaviours. Sleep analysis approaches are incorporated to increase the reliability of the system. This assessment is strengthened by utilising sleep stage detection and sleep position recognition. This includes, first, the extraction and integration of influence factors of sleep stage recognition methods to create a fine-grained personalised approach and, second, the detection of common but more complex sleep positions, including leg positions. The relations between medical conditions and sleep are assessed through interviews with doctors and users on various topics, including treatment satisfaction and technology acceptance. The findings from these interviews led to the investigation of sleep behaviour as a diagnostic indicator. Changes in sleep behaviour are assessed alongside medical knowledge using data mining techniques to extract information about disease development; the following diseases were of interest: sleep apnoea, hypertension, diabetes, and chronic kidney disease. The proposed framework is designed in a way that allows it to be integrated into existing smart home environments. We believe that our framework provides promising building blocks for reliable sleep behaviour assessment by incorporating newly developed sleep analysis approaches. These approaches include a modular layered sleep behaviour assessment framework, a sleep regularity algorithm, a user-dependent visualisation concept, a higher-granularity sleep position analysis approach, a fine-grained sleep stage detection approach, a personalised sleep parameter extraction process, in-depth understanding on sleep and chronic disease relations, and a sleep-wake behaviour-based chronic disease detection method.
- PhD