--------------------------------------------------- Date: Thu, 16 Sep 1999 09:35:35 +0200 From: Nils Olav =?iso-8859-1?Q?Sel=E5sdal?= To: PICLIST@MITVMA.MIT.EDU We're working at a project that uses 2 sound sensors do detect the angle to the sound source,like a load clap.The idea is to calculate(pic16f84??) this angle based on the time delay of the two sensors(placed about 10cm apart). I'm thankfull for any ideas, but i'm especially interrested in how-to-make the sound sensors. [nOs] --------------------------------------------------- Date: Thu, 16 Sep 1999 10:40:38 +0200 From: Dag Bakken Subject: Re: Sound directions To: PICLIST@MITVMA.MIT.EDU Something to simulate the human ear should do fine. The human ear are basically two omni-directional microphones placed at a fixed distance with sound-absorbant material inbetween. With this configuration, the incoming angle (not counting front/back/up/down) should be quite easy to calculate. The sound-absorbant material should be well-defined, so that you can make a fairly okay estimate on the sound caracteristic. If you require to read front and back as well, an imitation of the human ear's function should be quite possible (but not necessarily easy). The main function here is to alter the sound caracteristic based on the incoming angle. This is why the ear is shaped as it is. The problem with this approach is that the human brain "decodes" incoming sound by previously learned caracteristics. If this is to be done in a controller, the incoming sound should/must have some kind of reference so the controller can calculate deviation from original signal. If you design your hardware with two omnidirectional mics, a well-defined sound-absorbant between them, and a sound-absorbant with another caracteristic behind the mics, you should be able to (from a referenced/known sound) calculate left/right/front/back. If you don't want to design that, you can buy one. They are standard products, and (directly translated from norwegian) they're called 'artificial heads'. It is also possible to buy headsets/earphones with this facility built into them. This would most likely be a good solution since they can be "fairly small". -DS The Queue Principle: The longer you wait in line, the greater the likelihood that you are standing in the wrong line. Thursday, September 16, 1999, 9:35:35 AM, you wrote: NOS> We're working at a project that uses 2 sound sensors do detect the angle NOS> to the sound source,like a load clap.The idea is to calculate(pic16f84??) NOS> this angle NOS> based on the time delay of the two sensors(placed about 10cm apart). NOS> I'm thankfull for any ideas, but i'm especially interrested in how-to-make the NOS> sound sensors. NOS> [nOs] --------------------------------------------------- Date: Thu, 16 Sep 1999 09:49:09 -0400 From: Andy Kunz To: PICLIST@MITVMA.MIT.EDU >I'm thankfull for any ideas, but i'm especially interrested in how-to-make the >sound sensors. Nils, I would be interested in this as well. I did some work on an audio tachometer, and limited myself to shotgun mikes. To locate a direction, though, you'll need another technique, probably dealing as much with phase differential as anything else. Andy ================================================================== Eternity is only a heartbeat away - are you ready? Ask me how! ------------------------------------------------------------------ andy@rc-hydros.com http://www.rc-hydros.com - Race Boats andy@montanadesign.com http://www.montanadesign.com - Electronics ================================================================== --------------------------------------------------- Date: Thu, 16 Sep 1999 15:08:49 +0100 From: John Hallam Subject: Re: your mail To: PICLIST@MITVMA.MIT.EDU You only need this if you want to determine the direction of a once-off sound, though. If the sound has any reasonable duration (unlike the hand-clap signal mentioned earlier), you can do a fair job of finding its direction by swivelling your ears to maximise amplitude or swivelling your head to minimise arrival time difference. Reflections can throw this strategy off, but if the sensor system is moving the simple scheme works even better. (We built a phonotaxis system (modelled on the cricket) using a pair of tiny electret microphones, some electronic delay and mixing circuitry, and a couple of A/D's to sit on top of a Khepera mobile robot. It can follow the 4.7kHz mating calls emitted by male crickets. Unfortunately it can't mate with them ;-) Oh, and the programmable gains and delays were controlled by a PIC speaking I2C protocol to a variety of other bits and pieces.) John Hallam School of AI, Division of Informatics, University of Edinburgh, Scotland. On Thu, 16 Sep 1999, Andy Kunz wrote: > Nils, > > I would be interested in this as well. I did some work on an audio > tachometer, and limited myself to shotgun mikes. To locate a direction, > though, you'll need another technique, probably dealing as much with phase > differential as anything else. > > Andy --------------------------------------------------- Date: Thu, 16 Sep 1999 13:19:10 -0700 From: Mark Willis Subject: Locating objects using sound (was Re: (no subject)) To: PICLIST@MITVMA.MIT.EDU Andy Kunz wrote: > > >I'm thankfull for any ideas, but i'm especially interrested in how-to-make th e > >sound sensors. > > Nils, > > I would be interested in this as well. I did some work on an audio > tachometer, and limited myself to shotgun mikes. To locate a direction, > though, you'll need another technique, probably dealing as much with phase > differential as anything else. > > Andy Solving the equations for this, back in '82 or so, I got curves for time differentials between any 2 mikes. I think you need 3 mikes for 2-d or 3-d placement, I haven't done it in a while though Mark --------------------------------------------------- Date: Thu, 16 Sep 1999 13:05:35 -0400 From: "Robert A. LaBudde" Subject: =?iso-8859-1?Q?Re:_=99Check=5FSubject?= To: PICLIST@MITVMA.MIT.EDU At 09:35 AM 9/16/99 +0200, you wrote: >We're working at a project that uses 2 sound sensors do detect the angle >to the sound source,like a load clap.The idea is to calculate(pic16f84??) >this angle >based on the time delay of the two sensors(placed about 10cm apart). > >I'm thankfull for any ideas, but i'm especially interrested in how-to-make the >sound sensors. > >[nOs] Regular old electret mikes will work fine. They must be spaced a known distance apart. You need a set of 3 microphones to get the 2-D geometry correct. So long as the spacing between microphones is 10x the diameter of the mike aperture, the error is nil. Otherwise you can calibrate it out. ================================================================ Robert A. LaBudde, PhD, PAS, Dpl. ACAFS e-mail: ral@lcfltd.com Least Cost Formulations, Ltd. URL: http://lcfltd.com/ 824 Timberlake Drive Tel: 757-467-0954 Virginia Beach, VA 23464-3239 Fax: 757-467-2947 "Vere scire est per causae scire" ================================================================ --------------------------------------------------- Date: Thu, 16 Sep 1999 09:18:42 +0100 From: Michael Rigby-Jones Subject: Sound source angle detection To: PICLIST@MITVMA.MIT.EDU > We're working at a project that uses 2 sound sensors do detect the angle > to the sound source,like a load clap.The idea is to calculate(pic16f84??) > this angle > based on the time delay of the two sensors(placed about 10cm apart). > > I'm thankfull for any ideas, but i'm especially interrested in how-to-make > the > sound sensors. > > [nOs] > You don't have to make them...use microphones :o) Seriously it depends on what type of sounds you want to detect. Will it be any loud sound? If that's the case you will need the microphone, an active recitifier to amplify and rectify the AC signal from the microphone, and a comparator to give a clean logic signal to the PIC. A possible problem to watch out for is the possibility of reflections upsetting the calculations. Regards Mike Rigby-Jones --------------------------------------------------- Date: Thu, 16 Sep 1999 11:15:29 +0200 From: Dag Bakken Subject: Re: Sound source angle detection To: PICLIST@MITVMA.MIT.EDU MRJ> A possible problem to watch out for is the possibility of reflections MRJ> upsetting the calculations. That should'nt be a problem since a reflection would typically appear later than the time the sound uses from one mic to the other. The hazard being of course when the reflection comes from something closer to one mic than the distanse between them (the mics). -DS --------------------------------------------------- Date: Thu, 16 Sep 1999 02:28:00 -0700 From: Lynx {Glenn Jones} Subject: Re: Sound source angle detection To: PICLIST@MITVMA.MIT.EDU I have a question about this whole concept. Are you detecting sounds which are a fixed distance away, in this case i have no problem. However, how will you be able to detect the angle of the source when you have no distance information? ------------------------------------------------------------------------------ A member of the PI-100 Club: 3.1415926535897932384626433832795028841971693993751 058209749445923078164062862089986280348253421170679 On Thu, 16 Sep 1999, Dag Bakken wrote: > MRJ> A possible problem to watch out for is the possibility of reflections > MRJ> upsetting the calculations. > > That should'nt be a problem since a reflection would typically appear > later than the time the sound uses from one mic to the other. The > hazard being of course when the reflection comes from something closer > to one mic than the distanse between them (the mics). > > -DS > --------------------------------------------------- Date: Thu, 16 Sep 1999 11:54:42 +0200 From: Dag Bakken Subject: Re: Sound source angle detection To: PICLIST@MITVMA.MIT.EDU LGJ> I have a question about this whole concept. Are you detecting LGJ> sounds which are a fixed distance away, in this case i have no LGJ> problem. However, how will you be able to detect the angle of the LGJ> source when you have no distance information? Well... An angle/direction doesn't need distance information. The parameter that produces a center reading will be when there are no time-difference between left and right mic. This is of course true regardless of distance. The far left will be when the left mic reads the sound first, and the right mic read the same sound after the time it took the sound to travel. If the sound is somewhere in the front-left area, the sound will reach the right mic slightly earlier than when the source is at the far left. -DS MRJ> A possible problem to watch out for is the possibility of reflections MRJ> upsetting the calculations. >> That should'nt be a problem since a reflection would typically appear >> later than the time the sound uses from one mic to the other. The >> hazard being of course when the reflection comes from something closer >> to one mic than the distanse between them (the mics). --------------------------------------------------- Date: Thu, 16 Sep 1999 11:38:33 +0100 From: John Hallam Subject: Re: Sound source angle detection To: PICLIST@MITVMA.MIT.EDU On Thu, 16 Sep 1999, Dag Bakken wrote: > LGJ> I have a question about this whole concept. Are you detecting > LGJ> sounds which are a fixed distance away, in this case i have no > LGJ> problem. However, how will you be able to detect the angle of the > LGJ> source when you have no distance information? > > Well... An angle/direction doesn't need distance information. > [ ... snip ... ] This is almost right. The time difference measured at the microphones is directly related to the difference in path length between source -> left microphone and source -> right microphone. In 2 dimensions, the set of points whose distances from two fixed points (foci) differ by a constant is a hyperbola; in 3 dimensions, the surface you need results from spinning the hyperbola around the axis of the sensor system. What people normally do is (a) ignore elevation and assume the sound source is in the azimuthal plane and (b) assume the source is far enough away that the hyperbola looks like a straight line. For many applications these two approximations, while strictly false, are perfectly acceptable. John Hallam. Senior Lecturer, School of Artificial Intelligence, Division of Informatics, University of Edinburgh, Scotland. --------------------------------------------------- Date: Thu, 16 Sep 1999 03:42:58 -0700 From: Lynx {Glenn Jones} Subject: Re: Sound source angle detection To: PICLIST@MITVMA.MIT.EDU Thank you for the clarification. it was the Hyperbolic nature that was getting me, but your right that hyperbolas look pretty straight for this application. ------------------------------------------------------------------------------ A member of the PI-100 Club: 3.1415926535897932384626433832795028841971693993751 058209749445923078164062862089986280348253421170679 On Thu, 16 Sep 1999, John Hallam wrote: > On Thu, 16 Sep 1999, Dag Bakken wrote: > > > LGJ> I have a question about this whole concept. Are you detecting > > LGJ> sounds which are a fixed distance away, in this case i have no > > LGJ> problem. However, how will you be able to detect the angle of the > > LGJ> source when you have no distance information? > > > > Well... An angle/direction doesn't need distance information. > > [ ... snip ... ] > > This is almost right. The time difference measured at the > microphones is directly related to the difference in path length between > source -> left microphone and source -> right microphone. In 2 > dimensions, the set of points whose distances from two fixed points (foci) > differ by a constant is a hyperbola; in 3 dimensions, the surface you > need results from spinning the hyperbola around the axis of the sensor > system. > > What people normally do is (a) ignore elevation and assume the > sound source is in the azimuthal plane and (b) assume the source is far > enough away that the hyperbola looks like a straight line. For many > applications these two approximations, while strictly false, are perfectly > acceptable. > > John Hallam. > Senior Lecturer, School of Artificial Intelligence, > Division of Informatics, University of Edinburgh, > Scotland. > --------------------------------------------------- Date: Thu, 16 Sep 1999 13:18:33 +0200 From: Dag Bakken Subject: Re: Sound source angle detection To: PICLIST@MITVMA.MIT.EDU >> LGJ> I have a question about this whole concept. Are you detecting >> LGJ> sounds which are a fixed distance away, in this case i have no >> LGJ> problem. However, how will you be able to detect the angle of the >> LGJ> source when you have no distance information? >> >> Well... An angle/direction doesn't need distance information. >> [ ... snip ... ] JH> This is almost right. The time difference measured at the JH> microphones is directly related to the difference in path length JH> between source ->> left microphone and source -> right microphone. JH> In 2 dimensions, the set of points whose distances from two fixed JH> points (foci) differ by a constant is a hyperbola; in 3 JH> dimensions, the surface you need results from spinning the JH> hyperbola around the axis of the sensor system. Yes. So what you should do is measure elevation, right? What I explained in a (rather long) mail earlier today, was how the human brain/ears deals with front/back information. The same principle is used for elevation. I know that that approach is kind of far fetched for a controller, but it works. So...if that's implemented, you would effectively measure 3D - even though I think nothing less than AI should try it with that method. -DS --------------------------------------------------- Date: Thu, 16 Sep 1999 12:17:17 -0400 From: Wagner Lipnharski Subject: Re: Sound source angle detection To: PICLIST@MITVMA.MIT.EDU Suppose you can use two or more different tone frequencies at the emitter. In a direct sound wave, air mass can change the intensity level of some frequencies, based on the air sound transportability. Your receiver can identify this differences in a direct wave and store it in a non volatile memory. Different objects and surfaces will absorb and reflect different percentage of sound intensity. Only air can be compared to air. I believe no other object can be confused with air sound transportability. Even other gases have different transportability curves. If your receiver can compare those frequencies comparison received levels with the numbers stored from direct wave, it will be no only easy to identify a reflected wave, but also identify the reflector material. Suppose you install an obstacle (pole) exactly in the middle of the direct wave between the transmitter and receiver. The sound wave is a concentric wave that fight against itself resulting in a forward movement. The lack of wave after the pole will allow the wave to fill the gap and rebuild after the pole, of course with much less intensity than the direct wave. If you have a reflector to reflect waves to the receiver, this waves will reach the receiver stronger than the rebuilt direct waves. Your receiver could discriminate between both waves. The rebuilt wave will arrive at the receiver sooner than the reflected but with much less intensity. By this way it is possible to measure the shortest *air distance* between the origin and destination and the path via reflective waves. Remember that air still a transport vehicle to sound, and a non direct view path will always generate reflected or rebuilt waves, what could not be a direct line between origin and destination. A stereo rotational head can identify also the reflected angle by measuring the proportional distances between the angle of the rotational head. The comparison between the reflected and rebuilt waves can also helps to identify not only the reflector angle, but positional space of it. In real we do it all the time, and it is widely explored in surround systems, to create the illusion of 3 dimensional deep and distances. You can make this experiment; seat in a rotating chair, close your eyes and rotate the chair slowly. Keep a radio on at the other side of the room and ask somebody to walk in between you and the radio. Even with the chair rotating, you will be 100% able to not only identify when the person is obstructing the direct wave, but also guess the person distance from you, all based on reflected and rebuilt waves, echoes and different sound intensity levels for different frequencies. Wagner. --------------------------------------------------- Date: Thu, 16 Sep 1999 12:20:18 EDT From: Harold M Hallikainen Subject: Re: Sound source angle detection To: PICLIST@MITVMA.MIT.EDU On Thu, 16 Sep 1999 11:38:33 +0100 John Hallam writes: > > This is almost right. The time difference measured at the >microphones is directly related to the difference in path length >between >source -> left microphone and source -> right microphone. In 2 >dimensions, the set of points whose distances from two fixed points >(foci) >differ by a constant is a hyperbola; in 3 dimensions, the surface you >need results from spinning the hyperbola around the axis of the sensor >system. > > What people normally do is (a) ignore elevation and assume the >sound source is in the azimuthal plane and (b) assume the source is >far >enough away that the hyperbola looks like a straight line. For many >applications these two approximations, while strictly false, are >perfectly >acceptable. > Of course this hyperbola (just like the ones drawn on marine charts for LORAN navigation) go both in front and behind the person doing the listening. I've always wondered how we can tell if a sound is in front of us or behind us. I have two theories. 1. Due to the somewhat directional nature of our ears (they "point" forward), sounds behind us will have more reverberation than sounds in front of us, since there is more "gain" towards any reflector in front of us than towards the souce behind us. 2. On hearing a sound, we turn our head slightly. If we turn our head to the left and the sound got closer to the right ear (as determined by variation in time delay), the sound is in front of us. If it got farther from our right ear, the sound is behind us. I suspect that method 2 is the most probable. This would, however, make binaural headphones not sound natural, since the "image" would move when we turned our head. If binaural headphones had a direction sensor that fed back to a DSP of some sort that would insert the appropriate delays as we turned our head, we could perhaps get close to realistic directional effects in sound. Harold Harold Hallikainen harold@hallikainen.com Hallikainen & Friends, Inc. See the FCC Rules at http://hallikainen.com/FccRules and comments filed in LPFM proceeding at http://hallikainen.com/lpfm ___________________________________________________________________ Get the Internet just the way you want it. Free software, free e-mail, and free Internet access for a month! Try Juno Web: http://dl.www.juno.com/dynoget/tagj. --------------------------------------------------- Date: Thu, 16 Sep 1999 13:39:35 -0400 From: "Robert A. LaBudde" Subject: Re: Sound source angle detection To: PICLIST@MITVMA.MIT.EDU At 12:20 PM 9/16/99 -0400, Harold wrote: >On Thu, 16 Sep 1999 11:38:33 +0100 John Hallam >> This is almost right. The time difference measured at the >>microphones is directly related to the difference in path length >>between >>source -> left microphone and source -> right microphone. In 2 >>dimensions, the set of points whose distances from two fixed points >>(foci) >>differ by a constant is a hyperbola; in 3 dimensions, the surface you > Of course this hyperbola (just like the ones drawn on marine >charts for LORAN navigation) go both in front and behind the person doing >the listening. I've always wondered how we can tell if a sound is in It is my understanding that a sound generated from a single point would propagate by spherical (not hyperbolic) waves, which can be approximated by plane waves at long distances from the source. Since the speed of sound is nominally 340 m/s (1087 ft/s), you will have to time the differential pulse detection quite accurately to get reasonable accuracy from microphone arrays that are not spaced very far apart. ================================================================ Robert A. LaBudde, PhD, PAS, Dpl. ACAFS e-mail: ral@lcfltd.com Least Cost Formulations, Ltd. URL: http://lcfltd.com/ 824 Timberlake Drive Tel: 757-467-0954 Virginia Beach, VA 23464-3239 Fax: 757-467-2947 "Vere scire est per causae scire" ================================================================ --------------------------------------------------- Date: Fri, 17 Sep 1999 17:05:15 +0100 From: John Hallam Subject: Re: Sound source angle detection To: PICLIST@MITVMA.MIT.EDU A further thought on the sound direction discussion: why use only two microphones? If you want to discriminate azimuth and elevation a 3 microphone setup is simple and sufficient. You can also tell front/back with it by tilting the sensor plane, I think. (Or you can rely on the directional response of the microphones to reject sound from behind the sensor.) Biology is not always best... John Hallam, School of Artificial Intelligence University of Edinburgh --------------------------------------------------- Date: Fri, 17 Sep 1999 12:53:17 EDT From: Harold M Hallikainen Subject: Re: Sound source angle detection To: PICLIST@MITVMA.MIT.EDU On Thu, 16 Sep 1999 13:39:35 -0400 "Robert A. LaBudde" writes: > >It is my understanding that a sound generated from a single point >would >propagate by spherical (not hyperbolic) waves, which can be >approximated by >plane waves at long distances from the source. > True, I'd expect the sound to travel out in a sphere. However, the curve you get where the DIFFERENCE in distance to two points is a constant is a hyperbola. Let's see if I can remember it... I think it's 1 = x^2/a^2 - y^2/b^2 You get various conic sections by messing with this. The minus makes it a hyperbola. Change it to a plus and you get an ellipse where a and b are half the length of the axis. Make them equal and they are the radius of the circle. Harold Harold Hallikainen harold@hallikainen.com Hallikainen & Friends, Inc. See the FCC Rules at http://hallikainen.com/FccRules and comments filed in LPFM proceeding at http://hallikainen.com/lpfm ___________________________________________________________________ Get the Internet just the way you want it. Free software, free e-mail, and free Internet access for a month! Try Juno Web: http://dl.www.juno.com/dynoget/tagj. --------------------------------------------------- Date: Fri, 17 Sep 1999 15:38:00 -0700 From: Erik Reikes Subject: Re: Sound source angle detection To: PICLIST@MITVMA.MIT.EDU At 12:53 PM 9/17/99 -0400, you wrote: >On Thu, 16 Sep 1999 13:39:35 -0400 "Robert A. LaBudde" >writes: > >> >>It is my understanding that a sound generated from a single point >>would >>propagate by spherical (not hyperbolic) waves, which can be >>approximated by >>plane waves at long distances from the source. >> > > True, I'd expect the sound to travel out in a sphere. However, >the curve you get where the DIFFERENCE in distance to two points is a >constant is a hyperbola. Let's see if I can remember it... I think it's > > 1 = x^2/a^2 - y^2/b^2 > > You get various conic sections by messing with this. The minus >makes it a hyperbola. Change it to a plus and you get an ellipse where a >and b are half the length of the axis. Make them equal and they are the >radius of the circle. > Seems to me that givena reasonably sensitive receive circuit all you would have to do would be to have a directional detector, and slew it around on some kind of gimbal.... -Erik Reikes >Harold > > >Harold Hallikainen >harold@hallikainen.com >Hallikainen & Friends, Inc. >See the FCC Rules at http://hallikainen.com/FccRules and comments filed >in LPFM proceeding at http://hallikainen.com/lpfm > > >___________________________________________________________________ >Get the Internet just the way you want it. >Free software, free e-mail, and free Internet access for a month! >Try Juno Web: http://dl.www.juno.com/dynoget/tagj. --------------------------------------------------- Date: Sat, 18 Sep 1999 12:23:40 -0400 From: "Robert A. LaBudde" Subject: Re: Sound source angle detection To: PICLIST@MITVMA.MIT.EDU At 03:38 PM 9/17/99 -0700, Eric Reikes wrote: >>True, I'd expect the sound to travel out in a sphere. However, >>the curve you get where the DIFFERENCE in distance to two points is a >>constant is a hyperbola. Let's see if I can remember it... I think it's >> >> 1 = x^2/a^2 - y^2/b^2 >> >> You get various conic sections by messing with this. The minus >>makes it a hyperbola. Change it to a plus and you get an ellipse where a >>and b are half the length of the axis. Make them equal and they are the >>radius of the circle. >> > >Seems to me that givena reasonably sensitive receive circuit all you would >have to do would be to have a directional detector, and slew it around on >some kind of gimbal.... For some reason, I didn't get Harold's response to my post. The hyperbola is apparently the locus in the plane of all aliased points that would have given the same time delay at the dipole of microphone receivers. Obviously 2 mikes will not provide an accurate position in the plane, or even an angle. You need 3 mikes to do this. (This gives two independent timing intervals, which gives a single point locus.) With two intervals, the ability to time both (which may be simultaneous) accuracy is a performance issue. ================================================================ Robert A. LaBudde, PhD, PAS, Dpl. ACAFS e-mail: ral@lcfltd.com Least Cost Formulations, Ltd. URL: http://lcfltd.com/ 824 Timberlake Drive Tel: 757-467-0954 Virginia Beach, VA 23464-3239 Fax: 757-467-2947 "Vere scire est per causae scire" ================================================================ --------------------------------------------------- Date: Sat, 18 Sep 1999 13:10:12 -0400 From: "Sean H. Breheny" Subject: Re: Sound source angle detection To: PICLIST@MITVMA.MIT.EDU It seems to me that this situation is really analogous to light interference (or,perhaps closer in this case, arrays of RF receiving antennas). You can divide the situation into two regimes: the near field and the far field. In the near field, what you are saying is correct, the difference in path length is strongly linked to BOTH distance AND angle,so you cannot spearate the two and extract angle information with only two receivers(see note below). However,in the far field, the dependence on angle remains and the dependence on distance gets much less. If you keep the distance between the two mikes less than one wavelength, then there will be a one-to-one correspondence between physical angle and received phase difference,almost independent of changes in distance. If you are looking to measure sounds in the 1kHz range,this seems practical to me. However,it WOULD be difficult to do it for,say,20kHz,because your mike separation would have to be only about 1 cm,so the width of the mike would play a role in the phasing. It seems to me that someone else essentially said the same thing by saying that the hyperbola beomes flatter as you go farther out. Why was the idea dropped? NOTE: It seems to me that you might be able to play a trick and make this work in the near field,too. If you could determine the distance from the source to the mikes, you could extract the angle even in the near field. Since the sound level drops off with distance by a known relationship, and since you know the DIFFERENCE in distance (from the difference in received phase),you should be able to solve for the actual distance by looking at the difference in received signal strength between the two mikes). Even this would break down,though,if there were objects between source and mikes,OR if you brought the source so close that you were now in ITS near field,so that you no longer have the 1/r^2 dependence on sound level. Sean At 12:23 PM 9/18/99 -0400, you wrote: >For some reason, I didn't get Harold's response to my post. > >The hyperbola is apparently the locus in the plane of all aliased points >that would have given the same time delay at the dipole of microphone >receivers. > >Obviously 2 mikes will not provide an accurate position in the plane, or >even an angle. You need 3 mikes to do this. (This gives two independent >timing intervals, which gives a single point locus.) With two intervals, >the ability to time both (which may be simultaneous) accuracy is a >performance issue. > >================================================================ >Robert A. LaBudde, PhD, PAS, Dpl. ACAFS e-mail: ral@lcfltd.com >Least Cost Formulations, Ltd. URL: http://lcfltd.com/ >824 Timberlake Drive Tel: 757-467-0954 >Virginia Beach, VA 23464-3239 Fax: 757-467-2947 > >"Vere scire est per causae scire" >================================================================ > | | Sean Breheny | Amateur Radio Callsign: KA3YXM | Electrical Engineering Student \--------------=---------------- Save lives, please look at http://www.all.org Personal page: http://www.people.cornell.edu/pages/shb7 mailto:shb7@cornell.edu ICQ #: 3329174 --------------------------------------------------- Date: Fri, 17 Sep 1999 10:43:34 +1000 From: "W. Sierke" Subject: Re: Sound source angle detection To: PICLIST@MITVMA.MIT.EDU From: Harold M Hallikainen > Of course this hyperbola (just like the ones drawn on marine > charts for LORAN navigation) go both in front and behind the person doing > the listening. I've always wondered how we can tell if a sound is in > front of us or behind us. I have two theories. 1. Due to the somewhat > directional nature of our ears (they "point" forward), sounds behind us > will have more reverberation than sounds in front of us, since there is > more "gain" towards any reflector in front of us than towards the souce > behind us. 2. On hearing a sound, we turn our head slightly. If we I seem to recall being moderately impressed with an Aureal A3D-based sound card. One of their 3-D demos had a helicopter flying around you in a circular path. From recollection, as well as a pretty good front/rear impression, there was also some degree of altitude discernible, although this may have been most noticeable when the altitude was changing. (This was using stereo headphones, although the card also supported the use of front/rear speakers. I don't have good enough speakers to try it out with.) Wayne --------------------------------------------------- Date: Fri, 17 Sep 1999 08:31:58 +0200 From: Dag Bakken Subject: Re: Sound source angle detection To: PICLIST@MITVMA.MIT.EDU All direction of sound is mainly a product of two part. The most important one is learning. It has been tested that babies do not have the same stereo perseption as those a bit older. What they hear is more like two (similar) sounds. But, they learn oterwise. It's the same thing with direction, even though the mechanism is different. The brain learns how your ear picks up sound from different angles by learning how it should really sound, them how it actually sounds when placed above or behind or whatever. When the hearing mechanism has learned how it works, it can make assumptions when it hears a sound it hasn't heard before. That assumption will (most likely) not be as accurate, but close. Your ear actually changes the frequency spectrum of sound depending on direction. This is why an ear is shaped in the totally asymetric way it is. But since no ear are the same, it imposes quite a challenge for those recording Q-sound. No two people will hear the directions in a Q-sound recording the same way since their ears does not function excatly the same way, and may have learned slightly differently from what the Q-sound system is trying to reproduce. -DS HMH> the listening. I've always wondered how we can tell if a sound is in HMH> front of us or behind us. I have two theories. 1. Due to the somewhat HMH> directional nature of our ears (they "point" forward), sounds behind us HMH> will have more reverberation than sounds in front of us, since there is HMH> more "gain" towards any reflector in front of us than towards the souce HMH> behind us. 2. On hearing a sound, we turn our head slightly. If we HMH> turn our head to the left and the sound got closer to the right ear (as HMH> determined by variation in time delay), the sound is in front of us. If HMH> it got farther from our right ear, the sound is behind us. HMH> I suspect that method 2 is the most probable. This would, HMH> however, make binaural headphones not sound natural, since the "image" HMH> would move when we turned our head. If binaural headphones had a HMH> direction sensor that fed back to a DSP of some sort that would insert HMH> the appropriate delays as we turned our head, we could perhaps get close HMH> to realistic directional effects in sound. --------------------------------------------------- Date: Fri, 17 Sep 1999 13:41:32 +0200 From: Nils Olav =?iso-8859-1?Q?Sel=E5sdal?= Subject: Re: Sound source angle detection To: PICLIST@MITVMA.MIT.EDU charset="iso-8859-1" >We're working at a project that uses 2 sound sensors do detect the angle >to the sound source,like a load clap.The idea is to calculate(pic16f84??) this angle >based on the time delay of the two sensors(placed about 10cm apart). >I'm thankfull for any ideas, but i'm especially interrested in how-to-make the >sound sensors. we've gotten a bit further now, it wasn't all that bad. Yet it can't determin whether the sound is coming from the back or front. The sensors are placed 10cm apart, resolution of 9 degrees gives us about 70 instructions to detect the delay on a 10MHz 16f84.(a little optimizing will hopefully give us a resolution of 4.5 degrees.) Our method requires very little calculating. Just that every delay of whole 0.1/(340*10) are measured ,1 delay=9degrees! And as we use a resolution of 9 degrees, the failure CAN be big at great distances. [nOs]