Re: Optimal sweep duration for BRIR measurements (Bill Woods )

Subject: Re: Optimal sweep duration for BRIR measurements From: Bill Woods <Bill_Woods@xxxxxxxx> Date: Mon, 27 Jul 2015 15:49:38 +0000 List-Archive:<http://lists.mcgill.ca/scripts/wa.exe?LIST=AUDITORY> --_000_8B7E4506479CA84B8A00E10CF01B114301B94BDCEEep2pexmbs2mss_ Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable Hi All, Tough question from John, as it appears to have many dimensions - nonlinear= distortion, noise, time-varying noise, effort and resources. I have not st= udied the problem comprehensively but the thoughts below occurred to me. H= ope they're helpful. Cheers, Bill Woods 1. Philip brought up a good point, but technically speaking the sweep= does not have to be longer than the response, but the recording duration s= hould be. For example, in the ideal situation the impulse response of any r= oom could be measured with one impulse (i.e., a click). Just make sure you = record the whole thing. ;) 2. Speaking of clicks, I imagine the shorter the sweep, the more clic= k-like the stimulus. There are reasons sweeps are used instead of clicks (o= r MLS for that matter), so might be useful to consider them. Perhaps they u= nderlie the inaccuracies of the results with shorter sweeps? 3. I think Dick was speaking to the difficulties of working with the = large amount of data comprising recordings of very long signals, though tha= t would be odd to hear from someone at google. ;) Maybe I'm missing somet= hing, but clearly if there are several channels measured simultaneously at = a high sampling rate and long duration then the data storage, transfer, and= processing could become an issue if on a limited budget. 4. Speaking of long signals, I imagine the longer the sweep the more = you are stimulating the system with something like a sequence of long sinus= oids. I can't see any particular disadvantage of that relative to averaging= an equivalent number of responses obtained with moderate-length sweeps, an= d thus can't explain why your results with moderate-length sweeps are bette= r than with the longer. You mentioned the difficulties in the comparison of= results of the two lengths. Perhaps the averaging is not equivalent across= your moderate-length and long methods? Have you tried making a long record= ing in the room without a stimulus and using that recording in a simulation= of the two methods? At least you'd be able to compare them under the same = background noise conditions and check your implementation as well. Perhaps = there are fluctuations in the background noise that are not averaged-out si= milarly under the two methods? From: AUDITORY - Research in Auditory Perception [mailto:AUDITORY@xxxxxxxx= ILL.CA] On Behalf Of John Culling Sent: Friday, July 24, 2015 8:25 AM To: AUDITORY@xxxxxxxx Subject: Optimal sweep duration for BRIR measurements Dear all, Basic Q... Does anyone have insight into the optimum sweep duration using Farina's met= hod for measuring room impulses responses? More detailed background... We are planning to make an extensive series of measurements, and in prepara= tion have been testing the method using different sweep durations. One way to check t= he method is to correlate the impulses respones from repeated measurements or those g= enerated with different durations. To our surprise short sweeps (1-2 seconds) appea= r to give more reliable results (repeated sweeps correlate, r>0.98) than longer ones. Comp= aring sweeps of different durations is a little trickier, because we find a temporal off= set that reduces the correlation and can only be partially overcome by using cross-correlati= on. Nonetheless, it is apparent that durations from 1 second upwards correlate well, while g= oing below one second leads to reliable IRs, but ones that are inaccurate when compared wi= th those from longer sweep durations. Our surprising conclusion is that ~2s should be fine, but Farina refers to = an ISO standard that recommends very long sweeps (Farina has an example of 50s) to help overcome= noise. This seems an unintuitive rationale to us, since longer sweeps should incre= ase both the signal energy captured and the noise energy, and the method does not involv= e averaging as far as I understand. Longer durations should help address brief interupt= ing sounds, but I am unsure if that it what was the idea. In the presence of continuous noi= se, we did not notice any improvement in the IRs produced by longer sweeps. The nascent plan is to take >1 short sweep for each measurement and reject = IRs that that don't correlate well with another. Any insights/advice appreciated, John. Prof. John Culling School of Psychology, Cardiff University Tel: +44 (0)29 2087 4556 Yr Athro John Culling Yr Ysgol Seicoleg, Prifysgol Caerdydd Ff=F4n : +44 (0)29 2087 4556 --_000_8B7E4506479CA84B8A00E10CF01B114301B94BDCEEep2pexmbs2mss_ Content-Type: text/html; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable <html xmlns:v=3D"urn:schemas-microsoft-com:vml" xmlns:o=3D"urn:schemas-micr= osoft-com:office:office" xmlns:w=3D"urn:schemas-microsoft-com:office:word" = xmlns:m=3D"http://schemas.microsoft.com/office/2004/12/omml" xmlns=3D"http:= //www.w3.org/TR/REC-html40"> <head> <meta http-equiv=3D"Content-Type" content=3D"text/html; charset=3Diso-8859-= 1"> <meta name=3D"Generator" content=3D"Microsoft Word 15 (filtered medium)"> <style></style> </head> <body lang=3D"EN-US" link=3D"#0563C1" vlink=3D"#954F72"> <div class=3D"WordSection1"> Hi All,<o:p></o:p> <o:p> </o:p> Tough question from John, as it appears to have many= dimensions – nonlinear distortion, noise, time-varying noise, effort= and resources. I have not studied the problem comprehensively but the thou= ghts below occurred to me.  Hope they’re helpful.<o:p></o:p> <o:p> </o:p> Cheers,<o:p></o:p> Bill Woods<o:p></o:p> <o:p> </o:p> <![if !supportLists]>1.     &= nbsp; <![endif]>Philip brought up a good point, but technic= ally speaking the sweep does not have to be longer than the response, but t= he recording duration should be. For example, in the ideal situation the impulse response of any room could be measured = with one impulse (i.e., a click). Just make sure you record the whole thing= . ;)<o:p></o:p> <![if !supportLists]>2.     &= nbsp; <![endif]>Speaking of clicks, I imagine the shorter t= he sweep, the more click-like the stimulus. There are reasons sweeps are us= ed instead of clicks (or MLS for that matter), so might be useful to consider them. Perhaps they underlie the inaccuracie= s of the results with shorter sweeps?<o:p></o:p> <![if !supportLists]>3.     &= nbsp; <![endif]>I think Dick was speaking to the difficulti= es of working with the large amount of data comprising recordings of very l= ong signals, though that would be odd to hear from someone at google. ;)   Maybe I’m missing something, = but clearly if there are several channels measured simultaneously at a high= sampling rate and long duration then the data storage, transfer, and proce= ssing could become an issue if on a limited budget.<o:p></o:p> <![if !supportLists]>4.     &= nbsp; <![endif]>Speaking of long signals, I imagine the lon= ger the sweep the more you are stimulating the system with something like a= sequence of long sinusoids. I can’t see any particular disadvantage of that relative to averaging an equivalent number= of responses obtained with moderate-length sweeps, and thus can’t ex= plain why your results with moderate-length sweeps are better than with the= longer. You mentioned the difficulties in the comparison of results of the two lengths. Perhaps the averaging is = not equivalent across your moderate-length and long methods? Have you tried= making a long recording in the room without a stimulus and using that reco= rding in a simulation of the two methods? At least you’d be able to compare them under the same backg= round noise conditions and check your implementation as well. Perhaps there= are fluctuations in the background noise that are not averaged-out similar= ly under the two methods?<o:p></o:p> <o:p> </o:p> <o:p> </o:p></spa= n> <div> <div style=3D"border:none;border-top:solid #E1E1E1 1.0pt;padding:3.0pt 0in = 0in 0in"> <o:p> </o:p></spa= n> From: AUDITORY - Research in Auditory Percept= ion [mailto:AUDITORY@xxxxxxxx On Behalf Of John Culling Sent: Friday, July 24, 2015 8:25 AM To: AUDITORY@xxxxxxxx Subject: Optimal sweep duration for BRIR measurements<o:p></o:p> </div> </div> <o:p> </o:p> Dear all,<o:p></o:p> <o:p> </o:p> Basic Q…<o:p></o:p></span= > <o:p> </o:p> Does anyone have insight into t= he optimum sweep duration using Farina's method<o:p></o:p> for measuring room impulses res= ponses?<o:p></o:p> <o:p> </o:p> More detailed background…= <o:p></o:p> <o:p> </o:p> We are planning to make an exte= nsive series of measurements, and in preparation have<o:p></o:p> been testing the method using d= ifferent sweep durations. One way to check the method<o:p></o:p> is to correlate the impulses re= spones from repeated measurements or those generated <o:p></o:p> with different durations. &nbsp= ;To our surprise short sweeps (1-2 seconds) appear to give more <o:p></o:p> reliable results (repeated swee= ps correlate, r>0.98) than longer ones. Comparing sweeps<o:p></o:p></spa= n> of different durations is a lit= tle trickier, because we find a temporal offset that reduces<o:p></o:p></sp= an> the correlation and can only be= partially overcome by using cross-correlation. Nonetheless,<o:p></o:p></sp= an> it is apparent that durations f= rom 1 second upwards correlate well, while going below one<o:p></o:p></span= > second leads to reliable IRs, b= ut ones that are inaccurate when compared with those from<o:p></o:p>= longer sweep durations. <o:p></= o:p> <o:p> </o:p> Our surprising conclusion is th= at ~2s should be fine, but Farina refers to an ISO standard that <o:p></o:p> recommends very long sweeps (Fa= rina has an example of 50s) to help overcome noise.<o:p></o:p> This seems an unintuitive ratio= nale to us, since longer sweeps should increase both the <o:p></o:p> signal energy captured and the = noise energy, and the method does not involve averaging<o:p></o:p></= p> as far as I understand. Longer = durations should help address brief interupting sounds, but <o:p></o:p> I am unsure if that it what was= the idea. In the presence of continuous noise, we did not <o:p></o:p> notice any improvement in the I= Rs produced by longer sweeps.<o:p></o:p> <o:p> </o:p> The nascent plan is to take &gt= ;1 short sweep for each measurement and reject IRs that<o:p></o:p></= p> that don't correlate well with = another.<o:p></o:p> <o:p> </o:p> Any insights/advice appreciated= ,<o:p></o:p> <o:p> </o:p> John.<o:p></o:p> <o:p> </o:p> <table class=3D"MsoNormalTable" border=3D"0" cellspacing=3D"0" cellpadding= =3D"0" style=3D"border-collapse:collapse"> <tbody> <tr> <td width=3D"301" valign=3D"top" style=3D"width:225.4pt;border:solid #BFBFB= F 1.0pt;border-right:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt"> Prof. John Culling<o:p></o:p> School of Psychology,= Cardiff University<o:p></o:p> Tel: &#43= ;44 (0)29 2087 4556<o:p></o:p>= </td> <td width=3D"301" valign=3D"top" style=3D"width:225.4pt;border:solid #BFBFB= F 1.0pt;border-left:none;padding:0in 5.4pt 0in 5.4pt"> Yr Athro John Culling<o:p></o:p> Yr Ysgol Seicoleg, Pr= ifysgol Caerdydd <o:p></o:p> Ff=F4n : +44= (0)29 2087 4556<o:p></o:p> </td> </tr> </tbody> </table>  <o:p></o:p></s= pan> <o:p> </o:p> </div> </body> </html> --_000_8B7E4506479CA84B8A00E10CF01B114301B94BDCEEep2pexmbs2mss_--

This message came from the mail archive
/var/www/postings/2015/
maintained by:

DAn Ellis <dpwe@ee.columbia.edu>
Electrical Engineering Dept., Columbia University