Slashdot Mirror

← Back to Stories (view on slashdot.org)

UW Researchers Prototype Sonar-Based Contactless Sleep Monitoring

Posted by timothy on from the while-you-were-sleeping dept.

n01 writes: Researchers of the University of Washington are testing the prototype of their ApneaApp to diagnose sleep apnea, a health problem that can become life-threatening. To monitor a person's sleep, the app transforms the user's smartphone into an active sonar system that tracks tiny changes in a person's movements. The phone's speaker sends out inaudible sound waves, which bounce off a sleeping person's body and are picked back up by the phone's microphone. "It's similar to the way bats navigate," said Rajalakshmi Nandakumar, lead author and a doctoral candidate in the UW's department of computer science and engineering. "They send out sound signals that hit a target, and when those signals bounce back they know something is there." In technical terms, the app continuously analyzes changes in the acoustic room-transfer-function (sampled at ultrasonic frequencies) to detect motion. This is very similar to what the iPhone app Sleep Cycle Sonalarm Clock does, except that the UW researchers have improved the sensitivity of the method so it can precisely track the person's breathing movements which allows it to not only detect different sleep phases but also sleep apnea events. The advantage in both use cases is that the sleep monitoring is contact-less (there's nothing in the user's bed that could disturb their sleep) and doesn't require any additional hardware besides the user's smart phone.

24 of 40 comments (clear)

  1. Uh oh...Batman becomes real? by Shoten · · Score: 4, Interesting

    Turning smartphones into sonar devices to monitor movements. I'm torn between "this is really cool!" and "these people are so full of shit and just trying to publish something to get tenure!"

    I wonder how they solve the problems of directional discrimination without multiple microphones? How can they tell what direction a response comes from, with only one mic? And how do they intend to make this work on multiple phones, for that matter...with their vast differences in both microphone and speaker setups? I'm really skeptical of this.

    They also talk about using ultrasonic frequencies...which I also doubt most phones can actually produce.

    --

    For your security, this post has been encrypted with ROT-13, twice.
    1. Re:Uh oh...Batman becomes real? by Livius · · Score: 1

      How can they tell what direction a response comes from, with only one mic?

      It came from the person sleeping.

      The other problems, though, could be harder.

    2. Re:Uh oh...Batman becomes real? by n01 · · Score: 1

      I've read on a German website about the UW prototype that it requires a smartphone that can record from two microphones at the same time, so this probably solves the directional discrimination. The UW prototype uses 18-20 kHz which most adults can't hear. I know the iPhone's frequency range and it goes right up to 20 kHz for both playback and recording (disclaimer: I'm the developer of the Sonalarm app that I've referenced in the post and my app uses the 18.5 - 20 kHz range, IIRC).

    3. Re:Uh oh...Batman becomes real? by pushing-robot · · Score: 1

      Unless you're sleeping in a dormitory they don't need directionality; the phone can assume any fluctuation (doppler/impluse response/etc) comes from your movement.

      The phone's tiny speaker and mic can probably reach 20-24khz, which is *barely* ultrasonic but outside most human hearing. Your pets may no longer want to sleep with you, however.

      --
      How can I believe you when you tell me what I don't want to hear?
    4. Re:Uh oh...Batman becomes real? by Shoten · · Score: 2

      How can they tell what direction a response comes from, with only one mic?

      It came from the person sleeping.

      The other problems, though, could be harder.

      Which person? How can they tell the difference between the person sleeping and...

      The other person sleeping next to them?
      The pet in the room?
      Curtains, gently blowing in the breeze?
      The person shifting in their bed?
      Sounds from heating/cooling coming online and the air shifting around in the room as a result?

      How can it tell the difference between a response...a change in the state of something in the room...and a change in the object composition of the room itself? Without directionality, I don't see how it's possible. And indeed, as someone else pointed out, they did say that it requires phones with two microphones...which I missed when I read the article. So the point seems valid...and most phones won't be able to do this. Come to think of it, I am trying to think of what phones I know for a fact have dual microphones, and I'm coming up short.

      --

      For your security, this post has been encrypted with ROT-13, twice.
    5. Re:Uh oh...Batman becomes real? by Livius · · Score: 3, Insightful

      Most people don't bring their pets to medical examinations.

      A home test might have some utility but it's not a proper sleep study.

    6. Re:Uh oh...Batman becomes real? by n01 · · Score: 2

      I've had users report back that the Sonalarm app worked well for them while sharing the bed with their partner. You have a bit of directionality because both the loudspeaker and the mic are located at the bottom edge of the iPhone, and also range is limited to around 1 to 2 meters, depending on the selected sensitivity. (Disclaimer: I'm the developer of the Sleep Cycle Sonalarm Clock app that I've referenced in the post.)

    7. Re:Uh oh...Batman becomes real? by n01 · · Score: 1

      Starting with the iPhone 5, the iPhone actually has 3 built-in microphones. They are used to improve intelligibility during phone calls, but unfortunately an app can't record from multiple microphones directly (i.e. by getting 2- or 3-channel PCM sample data). I'm not sure how this is for Android phones.

    8. Re:Uh oh...Batman becomes real? by arielCo · · Score: 1

      Turning smartphones into sonar devices to monitor movements. I'm torn between "this is really cool!" and "these people are so full of shit and just trying to publish something to get tenure!"

      I wonder how they solve the problems of directional discrimination without multiple microphones? How can they tell what direction a response comes from, with only one mic?

      They don't, and they don't need to. Think of the ultrasound motion sensors in car / room alarms: if you emit chirps inside a closed volume they'll bounce off everything solid, and the pattern received at any point depends on everything inside, so you'll know if something moves. If you can keep track of the changes, you know if it's moving rhythmically and at what rate. Using multiple frequencies makes it more sensitive to changes, roughly speaking.

      And how do they intend to make this work on multiple phones, for that matter...with their vast differences in both microphone and speaker setups? I'm really skeptical of this.

      They also talk about using ultrasonic frequencies...which I also doubt most phones can actually produce.

      Again, no need. Put on some earplugs, or stick your head in a box, and you'll still recognize the beat of your favourite song in drastically altered acoustical conditions. The app is not measuring the transfer function to compare against some carefully calibrated curve, but the changes that tell it that something's moving, and with some smart processing it can tell apart your respiratory movements from the cat wandering in. A second person in the room might throw it off, though.

      They also talk about using ultrasonic frequencies...which I also doubt most phones can actually produce.

      This is the part that got me wondering. A cursory Google search gave me plots like this and this for the speaker and this one for the mic (yeah, condenser microphones have a pretty good range). So for this particular bit my answer is "feasible, and effectively inaudible if you're over 30".

      --
      This post contains no rudeness or derision of any kind. All arguments are friendly. Terms and exclusions may apply.
    9. Re:Uh oh...Batman becomes real? by Wescotte · · Score: 1

      Don't most modern phones have multiple microphones on them being used for noise reduction purposes?

    10. Re:Uh oh...Batman becomes real? by n01 · · Score: 1

      Starting with the iPhone 5, the iPhone actually has 3 built-in microphones. They are used to improve intelligibility during phone calls, but unfortunately on iOS an app can't record from multiple microphones directly (i.e. by getting 2- or 3-channel PCM sample data). I'm not sure how this is for Android phones. (Disclaimer: I'm the developer of the Sleep Cycle Sonalarm Clock app that I've referenced in the post.)

    11. Re:Uh oh...Batman becomes real? by roman_mir · · Score: 1

      Batman becomes real alright. Didn't you notice the acronym?

      UW - Universal Wayne.

      --
      You can't handle the truth.
    12. Re: Uh oh...Batman becomes real? by luisdom · · Score: 1

      Many phones have 2 mics, to do noise cancellation. Maybe it only works on those

    13. Re:Uh oh...Batman becomes real? by Bite+The+Pillow · · Score: 1

      It won't work well for people who can't sleep without their spouse or pet next to them, sure. But a controlled study won't do that person any good.

      Given my insurance, which is good, it would be cheaper to check in to a hotel and use this app to decide if I need formal diagnosis. By the time it comes out, speech recognition will require the hardware issues be solved, so no hardware expense.

      If I feel I have sleep issues and this says no apnea, I can look into other options before incurring medical tests.

      And they did test it in lab conditions, during actual sleep studies. But it is a lot simpler to secure your own room for one night, and may replace false positives from an unfamiliar environment, with false positives from a familiar one.

      Finally, they are apparently improving the state of the art in signal processing, so most of the readers here statistically don't know what that is, outside of similar reports. Meaning that some of your questions are solved problems, others are newly solved, and the rest being worked on. With a preliminary finding, this is more "neato, look what we found," instead of " we solved every problem."

    14. Re:Uh oh...Batman becomes real? by iluvcapra · · Score: 1

      How can it tell the difference between a response...a change in the state of something in the room...and a change in the object composition of the room itself? Without directionality, I don't see how it's possible.

      You create a baseline for the null or empty room, either by capturing the room empty or by creating an average of a few nights' observations.

      Transfer functions don't relate to spatial localization, they're different ways of figuring out directionality and both are independently reliable. Our heads use spatial localization (two ears) to detect the azimuth of a sound source, from both relative level and wavefront arrival time, but we determine the altitude of a sound source with transfer functions related to the shape of our heads and our outer ear structures.

      --
      Don't blame me, I voted for Baltar.
    15. Re:Uh oh...Batman becomes real? by n01 · · Score: 1

      I've heard back from the users of my app that Sonalarm actually worked pretty well when they shared their bed with their partner. The idea is that you place your iPhone on your bedside table (which has to be at your side of the bed, so the other person sleeping will be out of range).

    16. Re:Uh oh...Batman becomes real? by tlhIngan · · Score: 1

      Which person? How can they tell the difference between the person sleeping and...

      The other person sleeping next to them?
      The pet in the room?
      Curtains, gently blowing in the breeze?
      The person shifting in their bed?
      Sounds from heating/cooling coming online and the air shifting around in the room as a result?

      First off, sonar can be made discriminatory by simple correllation of the signal. That eliminates noises in the room by default. It's usually a brief chirp, and you only have to sample for a few milliseconds for a response.

      Second, we're not talking about a huge range here - someone puts their phone on the table and it really only has to scan up to around 4' (1.2m) for so. Scan any more and you'll have to deal with partners, pets, curtains, and the wall. This also means you can use a softer signal so you don't wake up the user.

      Partners, pets, etc aren't an issue when you're dealing with such a short range - you can assume the only hit you'll see is the person you're monitoring - and unless your partner or pet sleeps on top of you (which may explain your apnea problems...), well, it should work for basically anyone who needs it.

      And I 'm sure the doctor can ask the patient to not let their pets sleep on top of them, or to activate the unit when night-time bed activities are done.

  2. It is not going to work. by 140Mandak262Jamuna · · Score: 1

    The alarm clock that I need has to go the bathroom, fetch a glass of water and pour it on my face. Anything less is useless for me.

    --
    sed -e 's/Chuck Norris/Rajnikant/g' joke > fact
    1. Re:It is not going to work. by n01 · · Score: 1

      Good idea! Why don't you start a Kickstarter campaign for an alarm clock that pours water on your face to wake you up? No need for a robot to go to your bathroom - it can have a water tank.

  3. Sonar by maroberts · · Score: 1

    That should mean I can go to the bathroom without turning the lights on, just by having my phone emit ultrasound.

    --

    Donte Alistair Anderson Roberts - hi son!
    Karma: Chameleon

  4. Re:Why need sheep monitoring? by Anne+Thwacks · · Score: 1
    There is a possibility that Snowden shared data with sheep.

    (It is at least as plausible as many of the other stories linked to Snowden).

    --
    Sent from my ASR33 using ASCII
  5. Re:Great Solution, but ... by Hognoxious · · Score: 1

    For those of you who do not suffer from sleep Apena, hold your humorous comments for some other time and place.

    It could be worse, imagine being discyxlic.

    --
    Confucius say, "Find worm in apple - bad. Find half a worm - worse."
  6. Just listen for the snores by Anonymous Coward · · Score: 2, Interesting

    In all my nights doing polysomnographies, you rarely have silent obstructive sleep apnoeas, probably a better use for a microphone. If you have silent central types, you should worry about something else than dodgy sleep.

  7. kim toan thang by nguyenvanhai143 · · Score: 1

    http://www.kimtoanthang.vn/xa-...