Shop OBEX P1 Docs P2 Docs Learn Events
Hardware For Stingray To Detect Person Falling On Floor — Parallax Forums

Hardware For Stingray To Detect Person Falling On Floor

BobCochranBobCochran Posts: 8
edited 2009-12-18 18:05 in Robotics
Hi, I am thinking of buying the Stingray robot. I would like it to be able to listen for specific sounds such as: A person falling, or a specific person saying a specific word. I would like it to then move close to that person in response to the falling sound or the voiced word from that person. It should not respond to the same word spoken by others who might be in the vicinity. What additional hardware would I need to get the bot to do this much? Thanks! --Bob

Post Edited (BobCochran) : 12/6/2009 2:12:17 AM GMT

Comments

  • FranklinFranklin Posts: 4,747
    edited 2009-12-06 05:05
    I would develop the detection system first before buying the robot.

    ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
    - Stephen
  • Mike GreenMike Green Posts: 23,101
    edited 2009-12-06 06:12
    That's a very difficult problem. The SayIt module from Parallax can recognize a small number of words said by a specific speaker, but it's not that accurate. It may not be able to distinguish one speaker from another and it cannot tell where the sound is coming from. It's also not designed to detect spoken words at a distance. Echos and room noise very much disturb its accuracy. You might have a chance to do what you want by using an IR beacon pinned to the subject's clothing. You could also use a small radio remote control to signal an emergency. This could be made part of the same device.

    It's one thing to want to experiment with the concept and to make a crude prototype that sort of demonstrates under ideal conditions some of the ideas that you'd need for a real system. It's another thing to expect that you'd be able to actually build a practical system.
  • BobCochranBobCochran Posts: 8
    edited 2009-12-06 13:01
    Mike, thanks for your thoughts with this. I posed the question to get an idea of the difficulty level of what I'm asking. From what you are saying there isn't off-the-shelf technology that can enable the bot to do these things. And I had only listed the first of my ultimate goals. The whole idea is to have the bot monitor a quite sick person and identify situations where the person might need help. If help is needed then the bot should place phone calls and send text messages to a list of other people. I'd also like the bot to use a camera to capture and send video of the person to the same list of people. It looks like I have a real challenge ahead. Thanks! --Bob
  • Mike GreenMike Green Posts: 23,101
    edited 2009-12-06 15:15
    You're talking about something that, at a minimum, would need an embedded PC, probably running Linux, with a couple of hundred MB of RAM, maybe 8+GB of SD card storage, the equivalent of a sound card, a USB webcam, and WiFi. These are available, are quite compact (like the Gumstix), but you'd need some gel cells to power it all.
  • Ken GraceyKen Gracey Posts: 7,392
    edited 2009-12-06 16:56
    Hey Bob,

    I'd break the bigger project into many small pieces, addressing them one at a time. Some of the challenges you identified such as speech recognition (from a particular person) are way beyond anything I've seen done, at least with our microcontrollers. I notice that MSFT has some speaker-independent speech recognition built into their applications; maybe you need to throw a whole PC at the project for that part of the project.

    Ken Gracey
  • BobCochranBobCochran Posts: 8
    edited 2009-12-06 18:18
    Thanks for the feedback, Mike and Ken. I may well give this project a serious try because I can see some other uses for it too. I'll dub this project the NannyGrannyBot since the objective is to monitor a senior citizen -- my mother. I agree that I need to break the overall goal down into small pieces and address them one at a time. Thanks! --Bob
  • LeonLeon Posts: 7,620
    edited 2009-12-06 22:28
    You could use an accelerometer to detect falling. Image processing is another technique. It was one of the applications for the Irisys system that I helped to develop, but requires an IR image sensor in every room, and a lot of processing power.

    Leon

    ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
    Amateur radio callsign: G1HSM
  • BobCochranBobCochran Posts: 8
    edited 2009-12-06 22:46
    So far everyone agrees that I need a lot of processing power for what I want. What makes the most sense for me to start with, learning robotics first or learning an embedded PC platform like Gumstix first? I'm completely new to both of these platforms. I am pretty experienced with Linux -- been using it for about 10 years now. Mike's comments made me look up the Gumstix websites and begin looking at the products. Thanks! --Bob
  • sylvie369sylvie369 Posts: 1,622
    edited 2009-12-06 23:37
    Respectfully: I think that if this is a serious project you might want to rethink the basic method you're using. Unless you've already done quite a bit of this kind of thing, it's going to take you many months to come up with a version that does any of this, and many more months to make it moderately reliable.

    What if instead you had sort of an "opt-out" system, in which your mother had to check in regularly, and if she failed to do so in a certain period of time, then the messages were sent? Checking in could be simply a matter of touching an RFID tag to a reader once in a while. Augment it with that small radio remote that Mike mentioned, so that she can signal an emergency immediately, and you've got most of the functionality you're asking for, and you could have it in a matter of weeks rather than two or three years from now.

    Frankly, a robot - an actual moving device·like the Stingray - doesn't seem to me to offer any advantage, and it would require a LOT of tinkering (read: time and $) to make it even moderately reliable.

    Of course if the idea is to have some fun working with a Stingray, go for it. turn.gif
  • BobCochranBobCochran Posts: 8
    edited 2009-12-07 00:21
    An Opt-out system using an RFID is an interesting idea. I'll surely investigate and give it a lot of thought. I do like tinkering with small devices and have some electronics experience. I'm inexperienced with robotics but happy to learn and experiment. Thanks! --Bob
  • Phil Pilgrim (PhiPi)Phil Pilgrim (PhiPi) Posts: 23,514
    edited 2009-12-07 04:57
    It's important for an older person to feel that they still have some control over their life and don't have to be bothered constantly to answer a watchdog timer. My dad (now 92) used LifeLine for several years before entering assisted living, and my next-door neighbor (a lady in her 80s) has the system, too. Both have had positive results with it. In my dad's case, he received help immediately after a couple falls. Obviously, it won't be effective if the subject is unconscious, though. But I think you have to strike a balance between safety and intrusiveness.

    -Phil
  • BobCochranBobCochran Posts: 8
    edited 2009-12-07 22:07
    Well, to stick to the questions I've posed...I'm far more interested in exploring and developing this idea as a technical project. Thanks. --Bob

    Post Edited (BobCochran) : 12/7/2009 10:52:47 PM GMT
  • sylvie369sylvie369 Posts: 1,622
    edited 2009-12-08 12:23
    BobCochran said...
    Well, to stick to the questions I've posed...I'm far more interested in exploring and developing this idea as a technical project. Thanks. --Bob
    That's important to know, as this kind of project leads to as many questions about the "human factor" as it does about technical issues, as Phil pointed out. The basic question - how do I detect a problem with an ill person other than by self-report? - is a really interesting one.·A reliable, non-intrusive way of doing that would be VERY useful.
  • allanlane5allanlane5 Posts: 3,815
    edited 2009-12-08 17:13
    The big problem I've heard of is an older person tripping or falling, breaking a hip or a leg, and being so immobile they couldn't make it to a phone.

    In these cases, having an around-the-neck pendant of some description would be a good way to 'kick-off' the process.

    You really want some way to reduce "false positivies", where your system concludes the person is in trouble when in fact they're not. Emergency Personnel just hate responding to emergency requests when there's no emergency.
  • noobmunchernoobmuncher Posts: 124
    edited 2009-12-18 18:03
    I actually worked/ have been working on a similar project to this for a while now, so here is what I suggest.
    Build some sort of device that the user wears, for example a pendant as Allan suggests or a bracelet, belt whatever. It just can't be intrusive or awkward to wear. I would then install an accelerometer on it, and you could even install a heart rate monitor www.sparkfun.com/commerce/product_info.php?products_id=8660 (I am not sure if this one could be used, there is an interface device you can get for it as well). I would then have all of this controlled by a propeller chip and have one button on the outside of the device. If the person falls or any of the sensors detect something that you specify to mean a potential emergency the device would beep quite loudly and flash an LED. The wearer then has 10 seconds to press the exterior button otherwise it assumes it is an emergency. You could also have it so that the person could press and hold the button at any time in order to set off the same response.

    From here I would have the device send out some sort of signal, probably to a computer to send out the emergency protocol to wherever it needs to go. If you still think a robot is necessary consider having the device send out incredibly intense IR light to act like a beacon so the robot can find the persons location quite accurately.
  • noobmunchernoobmuncher Posts: 124
    edited 2009-12-18 18:05
    As far as battery life for the device goes I would have it be recharged each night or have swappable batteries or something.
    Good Luck! smile.gif
Sign In or Register to comment.