Sound Localization

[Commanderbob]

I wanted to know if anyone has tried to add sound localization (being able to detect which direction a sound came from) to a robot. I thought of doing it with three microphones in a triangle. An microcontroller could then detect a sound wave and measure how long it takes for it to get to the other two microphones. That way you could tell which direction it came from. I don't exactly know how well that would work or how you could one sound from another, but it seems like a good start to me. Just brainstorming.

Justin

[SeagullOne]

I'm plannig a similar design for my robot, NINA.

My theoretical approach is to have a ring of microphones attached to analogue-to-digital converters and measure how much voltage is received from which microphone. The microphone that returns the highest voltage is the most probable direction the sound comes from, or the robot can use a mathematical algorithm to find a "medium" in the voltage of the ring of microphones. The Robot can move toward the noise using a digital compass, encoders...whatever you see fit.

I don't know how well this would work either, though. Anyone Else?

[Commanderbob]

I thought about doing that only slightly different. Still using three microphones, some how filter their output so you get a smooth voltage and compare them with an ADC. I also think you would need to separate each microphone with some sort of barrier so sound from one direction affects it more than the others. Some thing like the inside of a peace sign.

[dunk]

i remember seeing a link once where someone was using a PC sound card with attached speaker and microphone for experimenting with sonar.
that might be a good way to research viable sound frequencies and microphone positions.

dunk.

[HDL_CinC_Dragon]

I dont think finding the maximum voltage would be helpful. The difference in sound level at that proximity would be immeasurably small. Your best bet would be to measure which mic spiked first. You would have to query the mics a hell of a lot each second though as sounds travels at around 340m/s.

[dunk]

ah, here it is:
http://eddiem.com/projects/chirp/chirp.htm
(sonar using your PC soundcard.)

dunk.

[SeagullOne]

That was one of my concerns, Dragon. You're right.

What about a ring of unidirectional microphones instead of omnidirectional microphones? This way the sound localization would only be received by microphones which are directly pointed at the source. Use an A2D converter for each microphone and find a medium.

My only concern with this idea would that the sound would have to be really close for the robot to localize it...

What do you guys think?

[fuzzyt]

You all have probably seen this, but it might be adapted to the localization problem.

http://www.societyofrobots.com/robotforum/index.php?topic=6038.0

[Ro-Bot-X]

I think you guys should read this page:

hghttp://www.convict.lu/Jeunes/Gaston/Gaston_Ears.htm

It has concept, schematics, program... and it was done in 2003. At the end, there is a link to the main page, you should also go through all of their pages, as their work is astonishing.

Better yet, look at the bottom of this page for the provided Links:

http://www.convict.lu/htm/rob/spatial_sound_localizing_robot_xyztuvw.htm

[SeagullOne]

Wow. Those methods look engaging to try!

[Kirk]

I have a "tape measure" that is an ultra sonic gadget. The transducer is at the bottom of a long tube. This makes the microphone extremely directional.  Perhaps such a microphone could be mounted on a hobby servo and could pan to listen for the beacon tone.  The loudest volume is the beacon. (or an echo)
Kirk

[Admin]

The http://www.auvsi.org competition has a section where robots are required to localize sound and move the robot to that sound.

Dozens of teams have done it, just search online for it.