Forums

Sound localisation

Started by And Blah Blah Blah February 1, 2006
I am doing research into sound localisation for a degree project and how 
to alter signals to affect their apparent position.  I understand how 
and why time delay and amplitude difference between the ears have an 
effect as well as spectral changes.  However I keep finding references 
to phase differences being an important cue and I don't understand what 
a phase difference would be in this context (why is it different from a 
time delay) and where it would come from.  Non of the references offer 
an explaination that I can understand so I wonder if anyone here can 
clear the matter up for me.

Thanks for any help

Richard
And Blah Blah Blah wrote:
> I am doing research into sound localisation for a degree project and how > to alter signals to affect their apparent position. I understand how > and why time delay and amplitude difference between the ears have an > effect as well as spectral changes. However I keep finding references > to phase differences being an important cue and I don't understand what > a phase difference would be in this context (why is it different from a > time delay) and where it would come from. Non of the references offer > an explaination that I can understand so I wonder if anyone here can > clear the matter up for me.
First, you might want to change the nickname that is displayed in the "name" field, to get more useful responses. Time differences and phased ifferences are basically the same thing. In frequency domain, shifting a signal in time corresponds to changing the phase. For practical purposes, a "phase shift" is a time delay tau that is smaller than the dominant period T, tau < T while a "time delay" is a time shift tau larger than the dominant period T, tau > T. HTH, Rune
Rune Allnor wrote:
> And Blah Blah Blah wrote: > >>I am doing research into sound localisation for a degree project and how >>to alter signals to affect their apparent position. I understand how >>and why time delay and amplitude difference between the ears have an >>effect as well as spectral changes. However I keep finding references >>to phase differences being an important cue and I don't understand what >>a phase difference would be in this context (why is it different from a >>time delay) and where it would come from. Non of the references offer >>an explaination that I can understand so I wonder if anyone here can >>clear the matter up for me. > > > First, you might want to change the nickname that is displayed in the > "name" field, to get more useful responses. > > Time differences and phased ifferences are basically the same thing. > In frequency domain, shifting a signal in time corresponds to changing > the phase. For practical purposes, a "phase shift" is a time delay tau > that is smaller than the dominant period T, > > tau < T > > while a "time delay" is a time shift tau larger than the dominant > period T, > > tau > T.
If bla bla bla delays an entire signal by a certain time, then each component will be shifted in phase by an amount proportional to its frequency. When a delay of a ms is created, 1KHz is retarded 360 degrees, 100 Hz 36, and DC not at all. (That's why we don't usually talk about the phase of DC. We can take it for granted.) Jerry -- Engineering is the art of making what you want from things you can get. &#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;
Jerry Avins wrote:

> Rune Allnor wrote: > >> And Blah Blah Blah wrote: >> >>> I am doing research into sound localisation for a degree project and how >>> to alter signals to affect their apparent position. I understand how >>> and why time delay and amplitude difference between the ears have an >>> effect as well as spectral changes. However I keep finding references >>> to phase differences being an important cue and I don't understand what >>> a phase difference would be in this context (why is it different from a >>> time delay) and where it would come from. Non of the references offer >>> an explaination that I can understand so I wonder if anyone here can >>> clear the matter up for me. >> >> >> >> First, you might want to change the nickname that is displayed in the >> "name" field, to get more useful responses. >> >> Time differences and phased ifferences are basically the same thing. >> In frequency domain, shifting a signal in time corresponds to changing >> the phase. For practical purposes, a "phase shift" is a time delay tau >> that is smaller than the dominant period T, >> >> tau < T >> >> while a "time delay" is a time shift tau larger than the dominant >> period T, >> tau > T. > > > If bla bla bla delays an entire signal by a certain time, then each > component will be shifted in phase by an amount proportional to its > frequency. When a delay of a ms is created, 1KHz is retarded 360 > degrees, 100 Hz 36, and DC not at all. (That's why we don't usually talk > about the phase of DC. We can take it for granted.) > > Jerry
But am I right in thinking that in the end they are not two seperate phenomonen? Whether the brain sees a shift in the whole signal by given time, or that each frequency has shifted in phase by a different but related angle; it is the same information expressed in a different form? RIchard
The Poster Formally Known As "And Blah Blah Blah" wrote:

   ...

> am I right in thinking that in the end they are not two seperate > phenomonen? Whether the brain sees a shift in the whole signal by given > time, or that each frequency has shifted in phase by a different but > related angle; it is the same information expressed in a different form?
Yes, but phase shift proportional to frequency -- so called "linear phase" -- isn't necessarily always the case. There are ways to shift the phase of all frequencies in a broad (but not limitless) band by an arbitrary and equal amount (90 degrees is most common), and so-called all-pass networks can shift phase vs. frequency almost arbitrarily without affecting amplitude response. A good description of what you really want to do or know will deter us (all right; me) from drowning you in unnecessary detail. Sometimes, I feel they "barfing up the book" is the only way I can avoid being misleading. When I know the point, I can stick to it. Jerry -- Engineering is the art of making what you want from things you can get. &#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;&#2013266095;
And Blah Blah Blah wrote:

> I am doing research into sound localisation for a degree project and how > to alter signals to affect their apparent position. I understand how > and why time delay and amplitude difference between the ears have an > effect as well as spectral changes. However I keep finding references > to phase differences being an important cue and I don't understand what > a phase difference would be in this context (why is it different from a > time delay) and where it would come from. Non of the references offer > an explaination that I can understand so I wonder if anyone here can > clear the matter up for me. > > Thanks for any help > > Richard
The references are probably to "interaural phase difference", in which case they relate to head-related transfer functions (HRTFs); and to the spectrum-modifying action of the pinnae. There is a short discussion of this in "3D sound" (Durand Begault 1994: p 57.). This is the salient quote: " HRTF Phase Charactersitics. There are also changes in interaural phase as a function of frequency that are caused by the HRTF. Phase delay as a function of frequency can be understood as follows: if a wideband sound source containing all audible frequencies were played instantaneously at the pinnae, some frequencies would arrive later than others. " So, your reference to spectral changes does incorporate aspects of IPD. The general point is that IPD varies with the individual, because of the unique shapes of individual pinnae, wherea ITD is a more basic measurement based solely on the distance between the ear, which many people can be assumed to share. So: phase difference is personal, time difference is general. You may find it useful to subscribe to the sursound list: (https://mail.music.vt.edu/mailman/listinfo/sursound) as there are many experts on localisation etc who can answer your question far better than I can (who is merely repeating what he has read!). Richard Dobson
in article LFcEf.13599$wl.2795@text.news.blueyonder.co.uk, Richard Dobson at
richarddobson@blueyonder.co.uk wrote on 02/01/2006 19:48:

> The references are probably to "interaural phase difference", in which case > they > relate to head-related transfer functions (HRTFs); and to the > spectrum-modifying > action of the pinnae. There is a short discussion of this in "3D sound" > (Durand > Begault 1994: p 57.). This is the salient quote: > > " > HRTF Phase Charactersitics. > > There are also changes in interaural phase as a function of frequency that are > caused by the HRTF. Phase delay as a function of frequency can be understood > as > follows: if a wideband sound source containing all audible frequencies were > played instantaneously at the pinnae, some frequencies would arrive later than > others. > " > > So, your reference to spectral changes does incorporate aspects of IPD. > The general point is that IPD varies with the individual, because of the > unique shapes of individual pinnae,
as depicted: http://www.donnapardue.com/earsonwallsmall.gif . this is possibly the main reason that pressing music for distribution doesn't do sound localization so good. you master it with an HRTF model in mind and the person listening has a different HRTF. but there are some things loosely in common. there are usually two prominent notches in the frequency response due to the pinnae and as the azimuth angle of the source increases the two notches increase roughly together (if i remember correctly) and as the zenith angle increases, the spacing between the two notches change (i think increase). -- r b-j rbj@audioimagination.com "Imagination is more important than knowledge."

robert bristow-johnson wrote:

> as depicted: http://www.donnapardue.com/earsonwallsmall.gif . this is > possibly the main reason that pressing music for distribution doesn't do > sound localization so good. you master it with an HRTF model in mind and > the person listening has a different HRTF. but there are some things > loosely in common. there are usually two prominent notches in the frequency > response due to the pinnae and as the azimuth angle of the source increases > the two notches increase roughly together (if i remember correctly) and as > the zenith angle increases, the spacing between the two notches change (i > think increase).
What I've found about pinnae differences and binaural reproduction, having done a good deal of it, is that while accuracy may suffer, plausibility doesn't. Subject X can listen to what was recorded with subject Y's ears and the result isn't any kind of scrambling or overall confusion. The listener generally accepts the 3-D image she hears as plausible with no real idea what deviations there might be from an accurate rendition. It doesn't seem that those deviations are gross enough to seem inappropriate or to take much away from the extraordinary experience that binaural offers. Many people with all sorts of pinnae have listened to what mine have done to impinging sound of considerable variety without once saying that it doesn't sound right. The universal reaction the first time is astonishment at the reality of the experience. Bob -- "Things should be described as simply as possible, but no simpler." A. Einstein
in article drs8e60nd6@enews1.newsguy.com, Bob Cain at
arcane@arcanemethods.com wrote on 02/02/2006 01:22:

> robert bristow-johnson wrote: > >> as depicted: http://www.donnapardue.com/earsonwallsmall.gif . this is >> possibly the main reason that pressing music for distribution doesn't do >> sound localization so good. you master it with an HRTF model in mind and >> the person listening has a different HRTF. but there are some things >> loosely in common. there are usually two prominent notches in the frequency >> response due to the pinnae and as the azimuth angle of the source increases >> the two notches increase roughly together (if i remember correctly) and as >> the zenith angle increases, the spacing between the two notches change (i >> think increase). > > What I've found about pinnae differences and binaural reproduction, > having done a good deal of it, is that while accuracy may suffer, > plausibility doesn't. Subject X can listen to what was recorded with > subject Y's ears and the result isn't any kind of scrambling or overall > confusion. The listener generally accepts the 3-D image she hears as > plausible with no real idea what deviations there might be from an > accurate rendition. It doesn't seem that those deviations are gross > enough to seem inappropriate or to take much away from the extraordinary > experience that binaural offers. > > Many people with all sorts of pinnae have listened to what mine have > done to impinging sound of considerable variety without once saying that > it doesn't sound right. The universal reaction the first time is > astonishment at the reality of the experience.
yeah, i can believe that. are there finally a published standardized set of HRTFs (and IAD) that can well be used for sound navigating music or movie sound? this would be a published set of frequency response curves (both magnitude and phase, i s'pose) as a function of different zenith and azimuth angles of the source. also, how sensitive is this from deviation from sitting in the "sweet spot"? -- r b-j rbj@audioimagination.com "Imagination is more important than knowledge."
robert bristow-johnson wrote:

..
> > yeah, i can believe that. are there finally a published standardized set of > HRTFs (and IAD) that can well be used for sound navigating music or movie > sound? this would be a published set of frequency response curves (both > magnitude and phase, i s'pose) as a function of different zenith and azimuth > angles of the source. also, how sensitive is this from deviation from > sitting in the "sweet spot"? >
The MIT set is old (based just on the KEMAR dummy-head), but compact: http://sound.media.mit.edu/KEMAR.html A much larger database including many pinna-specific ones, is at: http://interface.cipic.ucdavis.edu/CIL_html/CIL_HRTF_database.htm Of course, most HRTF-proceseed sounds are expected to be delivered to the listener over headphones, so the sweet spot does not apply. However, various cross-talk cancellation techniques )together with new speaker position paradims such as stereo dipoles) are used to enable good results over speakers. Generally, a google search on HRTF+pinna should find most of what is out there. Richard Dobson