A gadget, a position sensor in space, the minimum task is 'a person, or rather a head'?

Question

Sensors

The implementation is not critical, but the requirements are:
1. connecting the device to a standard computer (performance above average) using standard USB ports or at least COM, it is highly desirable to work not only in win (but support for win7 x64 is required), the presence of an SDK or open source tools to obtain coordinates
2. determination of the position relative to the 'base' minimum - in the form of 2 angles in three-dimensional space, maximum - in the form of 3 (you can have 2 angles and a distance, or 3-coordinates in the classical Euclidean space, it doesn't matter anymore, the question is already in the errors, the corresponding sizes of the working area and prices)
3. minimum working area: the distance from the sensor is at least 0.3 meters, maximum 4-5 meters, and at least 30-40 degrees vertically and horizontally
4. minimum response to triggering (delay between position change and data acquisition) - 30ms (30 signals per second)
5. minimum error - less than a centimeter within at least the first 2 meters of the working area, the inability to 'lose' the measured object (for example after blocking the object by another or leaving the working area), or rather, a quick automatic recovery of operability after such
failures reasonable battery life (3-5 hours or better days)
Optional:
7. simultaneous determination of the position of more than one object (for example, cheap transmitters and one base), at least two
8. the presence of at least one 'button' on the module, the coordinates of which must be measured (pressing is transmitted to the computer)
ps while we are talking about one copy, but if it is possible to get what you want in 'any' quantity, then it's great.
I can list some possible implementations of all this disgrace myself:
* 1-2 webcams (difficulty in implementing the 4th requirement - reaction speed) + marker on the object (bright color / simple geom. figure / ..) + software (on there were even articles with open source implementations somewhere)
If two webcams are at a distance but look at the same point, then it is quite easy to determine the distance to the object, you can also make one webcam but analyze the size of the desired object (less accurately and requires size setting)
Unfortunately, it is not so easy to find a video camera that is not overloaded with chips (read: paying extra reluctance), without a noticeable lag between movement and receiving information by the program. at the same time, the main lags give - a sensor (all cheap ones are guaranteed to lag in low light), software inside the camera (encoding), usb data transfer, decoding, image processing to detect an object. Just unrealistic requirements in 30ms.
Otherwise, this solution is only pluses and there are practically no problems in the 'knee-knee' implementation.
PRICE: HD cameras (they have a fairly fast sensor in VGA mode, some give out 60fps) they cost about 6-7t.r.
* video surveillance cameras - similar to a webcam, only pci / pci-e connection and no compression (by the way, try to find a simple device with a switchable mpeg codec), as an option - an analog 'video peephole' + tv tuner.
Encoding and decoding delays go away (and the analog eye also removes the sensor lag). Unfortunately, there is nothing to test on, is it possible with such an implementation to fit into 30ms for recognition, if anyone has the opportunity to check? it's very easy, display the image from the camera on the screen and move the resulting 'infinite tunnel', then measure the lag of the tunnel movement by the number of nested frames (for which nesting the delay from movement will be 1 second)
PRICE: surveillance camera from 1.5t.r. (video peephole from 500r), video editing fee from 1.5t.r. (you can buy a TV tuner for 700r) ... i.e. about 3-4t.r.
* a pair of video sensors connected directly to the board (the power of a simple atmel / pic is enough, I saw ready-made examples of detectors on the Internet), looking for a light bulb in the infrared range that blinks with a special signal.
High speed - there are sensors that allow you to shoot the selected 'window', which 'leads' the object under study.
Cons - it is unrealistic to find a ready-made solution; this is an option for those who are friends with a soldering iron and with straight hands
PRICE: ??? but about 3-5t.r.
* A set of (minimum three) ultrasonic distance detectors examining one object
Somewhere I even saw a usb version on sale, but some ugly prices will result (after all, you need at least three), as well as potential problems with their synchronous operation (i.e. you need to immediately look for a ready-made device, sharpened for calculating coordinates, and separate devices can interfere with each other), plus any extra object between or behind the sensors can introduce large errors (more precisely
, decoys) .
* Modification of the ultrasonic distance detector circuit - namely, a separate transmitter from the receiver (the transmitter is the one whose distance we determine, the signal sends simultaneously with the infrared signal, three ultrasonic receivers and one photodetector based on the difference in the receiving time between receiving light and sound signals)
I really hope for the presence ready-made solutions with a compact head-mounted sensor (in extreme cases, you can come up with something of your own)
PRICE: individual ultrasonic receivers and transmitters cost from 80r-150r, the circuit does not become more complicated than the previous one, the cost is theoretically about 1-2t.r.
Maybe there are other ideas? and / or ready-made links to online stores with delivery in Russia and good prices.

Answer 1

An IR LED is attached to the head, the position is determined by the Wii remote control.
Wiimote tracks the position of up to 4 IR sources, and sends ready-made coordinates via Bluetooth (as far as I remember, the camera resolution is 1024x768, the refresh rate is 100 Hz).
There are libraries in different languages ​​and for different OSes.
There is a proof of the concept from Johnny Chang Lee .

Answer 2

ps wii-remote and their Chinese counterparts are just too big, you can’t put them on your head, but the price of $ 12 says that the cost of the solution is a penny

Answer 3

video camera example: CamSpace example video , www.camspace.com/

Answer 4

About surveillance cameras: the delay on them, generally speaking, is minimal, but in order to work with an open SDK, for 5 tr, and even work, this does not happen, and they are not for that.
If you can attach LEDs (or at least markers) to your head, then two SD cameras + OpenCV may suffice (I'm not sure about 5 meters). I myself saw this on one camera built into the laptop and it worked.

Answer 5

Kinect. True price...

Answer 6

Look at TrackIR - I play Armu 2 with this one ... With a passive sensor, the radius is rather weak, but a kit with an active (pro) one costs about 200 dollars ($ 169.95 without delivery and customs clearance) ...