detecting humans

Tnadz

Hello all. I'm making a fairly large project which will be a sound chamber that has complete surround sound. I want it to be interactive as well. An example would be, there's a dissonant sound on one side of the chamber and if you move any part of your body towards it, it becomes more harmonious relative to the distance you are away from the sound. I've looked at several options but I can't find that perfect match. IR sensors can be finicky and won't detect ranges (at least the ones I've looked at), I can't use pings because they interfere with each other and are pretty poor at detecting people, and the chamber will be completely dark so I can't really use photoresistors. I've looked at hall effects/ Theramins for magnetic field detection but they also are tricky. I was hoping that there might be some other device that could detect the magnetic field that humans have around them. Also, the device doesn't have to be basic stamp as I will be using MAX/ MSP for most of the project functionality. But at this point I'll take what I can get. Thanks everyone for helping out.

~Ross

Jed

If you time the pings right you should be able to use the ping solution. They'd have to all be tied into the same microprocessor so the pings could be coordinated to not interfere with each other. I think you only need a pause of like 50ms for each ping's echos to subside. Lets say you've got 4 sensors (one in each direction), you could space each sensor's ping 50 ms from each other in a series and your overall ping delay for each sensor would be 200 ms. That's 5 pings a second. Would you need higher resolution than that?

EDIT: Actually you'd only need 2 ping sensors assuming the room is square. One on the N/S wall and one on the E/W wall. That would take your pings down to 10/second.

Post Edited (Jed) : 1/28/2008 9:48:26 PM GMT

David H.

If you assume that everyone will be sitting down, you could go "old school" and have a contact switch on each spring, cushion,... and have the processor deal with those sensors.

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
David


There are 10 types of people in this world,...
Those that understand binary numbers, and those that don't!!!

LilDi

I think the best way to detect humans in a room is with a PIR sensor. The sensor will only detect moving objects, but if you pan the sensor back and forth, it will detect a human, moving or not. The PIR sensor will ignore·all other objects in the room except maybe a heater or a bright light source. If you combine a PIR sensor and a Ping sensor panning together, the PIR will alert you of a human in the room and in what direction the person is with respect to the sensor, and the Ping will give you the distance the person is to the Ping sensor. Check out the links below. The Aconame PIR sensor is specifically designed to sense human heat signatures and ignore other heat sources such as heaters and bright light sources. Not sure of the specs on the Parallax PIR.

Parallax PIR:·· http://www.parallax.com/Store/Sensors/ObjectDetection/tabid/176/CategoryID/51/List/0/Level/a/ProductID/83/Default.aspx?SortField=ProductName%2cProductName

Acroname PIR:·· http://www.acroname.com/robotics/parts/R1-442-3.html

Franklin

How about video cameras and something like roborealm (or roborealm)

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
- Stephen

kelvin james

The pir scanner mentioned would be easiest way, i am not sure about any other electrical sensing that could be used, other than some type of pressure sensing network in the floor. If you want a higher resolution scanner for ir, as to detect actual body movement then i might have something for you to look at.

One of my projects on the backburner involves human detection / following. The idea is to use an ir temperature scanner for the sensor. I got the idea from my handheld, it can detect temperature differences over as little as a couple of inches within 15 ft. And it has a fairly quick response time, which would work for a scanner. Servo feedback would provide position location. The amount of scanners would depend on the accuracy needed, and the size of the room.

The problem is finding an oem module for a reasonable cost and simple interface. I found one at the link shown below, but only has a 1:1 lense which won't cut it for me, but may work for short distances. The spi interface could be a problem also.

Anyway, not a ready made solution or tested, but might be worth some research.


www.metrisinst.com/product-details?ID=42&sm_chart=0

Dgswaner

I think you need to look at Roborelm as Franklin suggested. You could use a IR camera for night vision, and with the right filters the software could easily give you location and distance data.

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
"A complex design is the sign of an inferior designer." - Jamie Hyneman, Myth Buster

DGSwaner

allanlane5

The 'ping' has the highest probability of working. "PIR" sensors will tell you if something has 'moved' in a room -- but they're not good at telling you exactly where in the room something is.

For the speed of response and sensitivity (like detecting a moving hand, and EXACTLY how near or far it is) the ping would be most practical.

LilDi

Dgswaner,
···· I have looked at thermal IR night vision cameras. I have to say, even the really cheap ones cost $1000.00.
Do you have a source for a cheaper one. I would be very interested.

Roborealm, is an interesting approach. Maybe the best solution. Correct me if I'm wrong, but Roborealm is object recognition software that runs on a PC.

There is a very new embedded object recognition hardware module from Recognetics that is showing great promise as well.

http://www.general-vision.com/publications/CogniMem_tech_intro.pdf

www.recognetics.com/download/DS_CM_EB1K.pdf

Dgswaner

I wouldn't say that it's object recognition, but for lack of a better term that might be what I would call it.

I wasn't suggesting that you use a a thermal IR camera. just a camera with night vision.

let me step back for a sec... roborealm has a filter that will block out every thing except skin tones, it can then give relative size of the skin toned area it recognizes, and based on the position in the image area, a distance/location. This is what I would recommended you use, if you weren't doing it in the dark, but in theory, you could create a "night vision skin filter" that would do the same thing.

This is further than I have been in roborelm but I believe that you could also compare a empty room picture with live video and have it determine whats different.....where the people are. so if you mounted a camera(s) on the ceiling it would be really easy to determine where people where.

Roborelm will send variable data via serial, so it's easy to receive info. I would just use an ordinary web cam in a lit room to get a feel for it. See if you think it will work in your application. They have a tutorial for people recognition.

as with most applications I think a combination of sensors will yield a greater accuracy.

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
"A complex design is the sign of an inferior designer." - Jamie Hyneman, Myth Buster

DGSwaner

Post Edited (Dgswaner) : 1/28/2008 5:11:37 PM GMT

kelvin james

Any ccd based camera should be able to see ir, unless they have an ir filter built into it. This is a link to hacking a webcam for ir use.

www.lpi.usra.edu/education/fieldtrips/2005/activities/ir_spectrum/ir_webcam.html


IR cameras are plentiful now for security, like this one with built in ir leds.

www.tigerdirect.com/applications/searchtools/item-details.asp?EdpNo=1857310

LilDi

Get a roll of Ektakrome color slide film.
Expose it to the light and get it developed. You will now have completely blackened slides.
Cut out a piece of film the same size as the plastic filter you removed and put it in its place.
You now have a camera with an IR pass filter that block visible light.
You get some really errie looking images.

Post Edited (LilDi) : 1/28/2008 6:09:34 PM GMT

Tnadz

WOW! My comp crashed over the weekend so i couldn't check up on things till now but i never expected to many responses. It makes me cry a little. I think I do want to go the way of the camera. The pings sensors work but because the room gives people room to move, there would be a lot of glancing on clothes (especially winter clothes). I actually thought of a series of break beams but I want to try and get one complex sensor. I know, it isn't the best idea, but in this case, it will allow me to do it in the time frame i have. This may be a long shot, but is there anything useful out there as far as interfacing with Jitter? Or even just a com port/usb on the PC or Mac? Again thank you for all the incredibly helpful info.

~Tnadz