We could combine the use of various equipment available in smart homes or even social networks to increase the accuracy of sleep monitoring. Depending on the focus, one could have a component-based solution to achieve fault tolerance (if it is the goal) or use embeddings of different models built with different sensors.

One possible question: How to improve accuracy in sleep monitoring through the use of embeddings?

Great question, Lukas! As you will be aware, sleep deprivation, including partial sleep deprivation, and in particular reducing the amount of REM sleep, has very rapid antidepressant effects. Conversely, one can make a good case that too much REM sleep may cause depressive symptoms such as fatigue, lack of energy, lack of motivation, and so on. Because REM sleep propensity is strongly controlled by where a person is in their circadian rhythm cycle, maintaining regular sleep patterns, especially the time of getting up, appears to be very important to maintain mood/affect stability.

Consequently, an accurate (and easy!) way for individuals to monitor both their sleep stages as well as their circadian rhythms, might become part of an algorithm to help individuals adjust their behaviour in order to control their mood state, even in people with bipolar disorder. This, of course, requires research to confirm the hypothesis and to optimize the algorithms.

If your organization is working on easy-to-use technology for sleep stage monitoring, I would be very interested in collaborating.

Felipe Taliar Giuntini, thank you for your answer! Smart homes are truly a potential integration suitable for a sleep monitor of the above mentioned quality. What type of sensors are you thinking about? We are using a close-range ultra-wideband ultra-low-power radar for movement detection, today.

Henry Olders, the role of REM sleep in mental health and mood stability is paramount. I am not completely aware of the research connected to «too much REM sleep», that you are referring to. Would you be able to share that with me?

Feel free to take contact through [email protected]. My organisation is interested in studying the effect and implications of individual sleep interventions to provide tailored sleep health advice. We can today measure sleep quality and circadian rhythm completely non-intrusive and highly accurate. I see great collaboration potential. Let me know!

Lukas Krondorf , Thank you for your interest. Please have a look at the fulltext of my article, "Average sunrise time predicts depression prevalence" which references the literature re excessive REM sleep and depression.Article Average sunrise time predicts depression prevalence

However, definitive testing of the hypothesis awaits suitable instrumentation for accurately determining circadian rhythms, sleep, and sleep stages for use by individuals at home, and requiring minimal intervention by the individual or the researcher. Perhaps your technology will enable that research.

A case can be made that some depressive symptoms such as agitation or decreased activity (as seen in retarded depressions) can be detected and quantified by smart home technology such as yours, or the wifi approaches eg https://news.mit.edu/2018/artificial-intelligence-senses-people-through-walls-0612. If one could at the same time identify which individual, for households with more than one person, these technologies could be useful not only for diagnosing affective disorder, but also to help select appropriate treatments and to track treatment effectiveness.

Lukas Krondorf I think about light, movement, and noise sensors in general. I believe it is interesting to explore existing home equipment. Works by colleagues and from what I know in the literature, fog-computing, and publish-subscribe paradigms have been used. So, think that you could have several applications in a smart home, including sleep monitoring. You could use all the architecture of the existing house (equipment such as Smarts TVs, lamps, refrigerators, and objects in general), both for sensing and for processing data. Issues of offloading and data fusion have also been explored in this context.

Henry Olders, this was very interesting research, you shared. I believe the instrumentation through our non-contact sleep sensor might be the missing piece to the puzzle of circadian rhythm disorder and depression. We have now published our validation study, which you might want to check out here: Article Validation of sleep stage classification using non-contact r...

Both you and Felipe Taliar Giuntini have correctly pointed out the importance of smart home sensors. I am not familiar with publish-subscribe paradigms. Would you be able to point me in the right direction of how to apply this in a smart home environment?

Do you think about the conundrum of personal data collection through home sensors and their coupling with research? Data protection must play a huge role in how to monitor patients vs. the average person.

Lukas Krondorf I share your concern re data protection, and I think I have an answer. I've developed a neural network based classifier tool which performs well and has a very tiny footprint in terms of memory and processing resources. It can run on someone's smartphone, or perhaps within your device, and as such it would not need any data to be uploaded to anywhere else.

Lukas Krondorf the paper ( http://downloads.hindawi.com/journals/wcmc/2019/2786837.pdf) of my friend José R. Torres Neto presents an architecture that uses Publish-subscribe paradigm in smart homes. The validation process is done with applications on the recognition of emotions.