Subject: Re: Reference for typical SNRs is public spaces From: Bill Woods <Bill_Woods@xxxxxxxx> Date: Fri, 24 Jan 2014 21:51:49 +0000 List-Archive:<http://lists.mcgill.ca/scripts/wa.exe?LIST=AUDITORY>--_000_8B7E4506479CA84B8A00E10CF01B1143D0FBF33Cep2pexmbs2mssta_ Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: quoted-printable Dear List, Since we're assuming Andy is asking about speech-to-noise ratio when referr= ing to SNR in public spaces, I wondered if the question couldn't be answere= d by back-of-the-envelope calculations using some assumptions and articulat= ion index theory (AI). It turned out there was quite a long list of assumptions behind the calcula= tion (see below), but the resultant simple calculation yielded a SNR of 0.5= dB, which is quite in line with the values other posters have indicated ma= y actually be present. Two other facts became apparent as I did this, howev= er, and further motivate my posting. First, the long list of assumptions makes apparent the long list of influen= ces on the SNR in such a situation. These influences need to be characteriz= ed in any attempt to generalize from SNR measurements in a given scenario, = a fact other posters have alluded to. Second, it is important to note that much work has been done recently to mo= del or empirically characterize these influences on speech intelligibility,= and that, while the models may require more computational power than found= on the back of an envelope, they are no problem to execute on current lapt= op computers. The implication is that it should not be very difficult to de= termine a reasonably-accurate distribution of SNRs over wide variation in t= he assumed listening scenario without any measurements. Cheers, Bill Bill Woods, PhD Principal Research Scientist | Starkey Hearing Research Center 2150 Shattuck Ave. | Suite 408 | Berkeley, CA 94704-1345 T: 510-845-4876 x 14 starkey.com<http://www.starkey.com/> | starkeyresearch.com<http://www.stark= eyresearch.com> | map<http://maps.google.com/maps?q=3D2150+Shattuck+Avenue,= +Berkeley,+CA+94704&hl=3Den&ll=3D37.869941,-122.268219&spn=3D0.028152,0.038= 066&sll=3D37.86923,-122.273197&sspn=3D0.056305,0.076132&z=3D15> | email<mai= lto:william_woods@xxxxxxxx> The assumptions are: 1. Talkers have no hearing loss and no cognitive loss. 2. The talkers are facing each other and speaking in their "mother ton= gue". 3. We know the percent-correct (%C) targeted by the talkers and it is = less than 100% (it's a challenging environment). 4. We know the nature of the speech in such a conversation, from a low= -context vs. high-context perspective. 5. The talkers are within their critical distance (i.e., ignore reverb= eration of talkers' speech). 6. We know the long-term spectral shape of speech at the eardrums. 7. We know the long-term spectral shape of noise at the eardrums. 8. The noise is stationary. 9. The noise is diffuse. 10. The diffuse noise in combination with diotic direct-wave target spe= ech generates the equivalent of an "internal" wideband binaural SNR improve= ment of ~1.0 dB over monaural listening. 11. Overall level is not too high (i.e., no "roll-over" effect for inte= lligibility has occurred). 12. We're not including lip reading. These assumptions allow one to, first, determine the AI needed to achieve t= he assumed target %C given the assumed type of speech, and, second, determi= ne the SNR required with the assumed spectral shapes to obtain that AI. For instance, assuming talkers want 95%C with "unfamiliar sentences" then u= sing the polynomial fits from Sherbecoe and Studebaker (JASA 1990) of the A= NSI S3.5-1969 transfer functions between AI and %C, our talkers would need = an AI of 0.45. If we assume the noise and speech have the same long-term sp= ectral shape then the SNR can be determined from (SNR+12)/30=3D0.45 (stayin= g with the 1969 AI method), yielding SNR =3D 1.5 dB. Subtracting the binau= ral SNR improvement yields 0.5 dB. From: AUDITORY - Research in Auditory Perception [mailto:AUDITORY@xxxxxxxx= ILL.CA] On Behalf Of Andy Sabin Sent: Wednesday, January 22, 2014 9:53 AM To: AUDITORY@xxxxxxxx Subject: Reference for typical SNRs is public spaces Hi List, Can anyone point me to a reference showing SNRs that are typically observed= in public spaces (e.g., restaurants, bars ...etc)? I can find this info fo= r overall SPL, but am having a hard time finding it for SNR. Thanks Andy Sabin --_000_8B7E4506479CA84B8A00E10CF01B1143D0FBF33Cep2pexmbs2mssta_ Content-Type: text/html; charset="us-ascii" Content-Transfer-Encoding: quoted-printable <html> <head> <meta http-equiv=3D"Content-Type" content=3D"text/html; charset=3Dus-ascii"= > <meta name=3D"Generator" content=3D"Microsoft Exchange Server"> <!-- converted from rtf --> <style><!-- .EmailQuote { margin-left: 1pt; padding-left: 4pt; border-left:= #800000 2px solid; } --></style> </head> <body> <font face=3D"Calibri" size=3D"3"><span style=3D"font-size:12pt;"> <div>Dear List,</div> <div><font face=3D"Times New Roman"> </font></div> <div>Since we’re assuming Andy is asking about speech-to-noise ratio = when referring to SNR in public spaces, I wondered if the question couldn&#= 8217;t be answered by back-of-the-envelope calculations using some assumpti= ons and articulation index theory (AI). </div> <div> </div> <div>It turned out there was quite a long list of assumptions behind the ca= lculation (see below), but the resultant simple calculation yielded a SNR o= f 0.5 dB, which is quite in line with the values other posters have indicat= ed may actually be present. Two other facts became apparent as I did this, however, and further motivate my= posting. </div> <div> </div> <div>First, the long list of assumptions makes apparent the long list of in= fluences on the SNR in such a situation. These influences need to be charac= terized in any attempt to generalize from SNR measurements in a given scena= rio, a fact other posters have alluded to.</div> <div> </div> <div style=3D"margin-bottom:12pt;">Second, it is important to note that muc= h work has been done recently to model or empirically characterize these in= fluences on speech intelligibility, and that, while the models may require = more computational power than found on the back of an envelope, they are no problem to execute on current lapto= p computers. The implication is that it should not be very difficult to det= ermine a reasonably-accurate distribution of SNRs over wide variation in th= e assumed listening scenario without any measurements. </div> <div>Cheers,</div> <div>Bill</div> <div><font face=3D"Times New Roman"> </font></div> <table width=3D"471" style=3D"width:282.75pt;margin-left:5.4pt;"> <col width=3D"471" style=3D"width:282.75pt;"> <tr> <td><font face=3D"Arial" color=3D"#17365D">Bill Woods, PhD<br> <font size=3D"2" color=3D"#595959"><span style=3D"font-size:10pt;">Principa= l Research Scientist | Starkey Hearing Research Center<br> 2150 Shattuck Ave. | Suite 408 | Berkeley, CA = ;94704-1345<br> T: 510-845-4876 x 14 </span></font></font></td> </tr> <tr> <td><font face=3D"Times New Roman"><a href=3D"http://www.starkey.com/"><fon= t face=3D"Arial" size=3D"2" color=3D"blue"><span style=3D"font-size:10pt;">= <u>starkey.com</u></span></font></a><font face=3D"Arial" size=3D"2" color= =3D"#0000F6"><span style=3D"font-size:10pt;"> </span></font><font face= =3D"Arial" size=3D"2" color=3D"#595959"><span style=3D"font-size:10pt;">|</= span></font><font face=3D"Arial" size=3D"2" color=3D"#0000F6"><span style= =3D"font-size:10pt;"> </span></font><a href=3D"http://www.starkeyresea= rch.com"><font face=3D"Arial" size=3D"2" color=3D"blue"><span style=3D"font= -size:10pt;"><u>starkeyresearch.com</u></span></font></a><font face=3D"Aria= l" size=3D"2" color=3D"#0000F6"><span style=3D"font-size:10pt;"> </spa= n></font><font face=3D"Arial" size=3D"2" color=3D"#595959"><span style=3D"f= ont-size:10pt;">|</span></font><font face=3D"Arial" size=3D"2" color=3D"#00= 00F6"><span style=3D"font-size:10pt;"> </span></font><a href=3D"http:/= /maps.google.com/maps?q=3D2150+Shattuck+Avenue,+Berkeley,+C= A+94704&hl=3Den&ll=3D37.869941,-122.268219&spn=3D0.028152,0= .038066&sll=3D37.86923,-122.273197&sspn=3D0.056305,0.076132&z= =3D15"><font face=3D"Arial" size=3D"2" color=3D"blue"><span style=3D"font-s= ize:10pt;"><u>map</u></span></font></a><font face=3D"Arial" size=3D"2" colo= r=3D"#0000F6"><span style=3D"font-size:10pt;"> </span></font><font fac= e=3D"Arial" size=3D"2" color=3D"#595959"><span style=3D"font-size:10pt;">|<= /span></font><font face=3D"Arial" size=3D"2" color=3D"#0000F6"><span style= =3D"font-size:10pt;"> </span></font><a href=3D"mailto:william_woods@xxxxxxxx= arkey.com"><font face=3D"Arial" size=3D"2" color=3D"blue"><span style=3D"fo= nt-size:10pt;"><u>email</u></span></font></a></font></td> </tr> <tr> <td><font face=3D"Times New Roman"></font></td> </tr> <tr> <td><font face=3D"Times New Roman"></font></td> </tr> </table> <div><font face=3D"Times New Roman"> </font></div> <div>The assumptions are:</div> <ol style=3D"margin:0;padding-left:36pt;"> <li>Talkers have no hearing loss and no cognitive loss.</li><li>The talkers= are facing each other and speaking in their “mother tongue”.</= li><li>We know the percent-correct (%C) targeted by the talkers and it is l= ess than 100% (it’s a challenging environment).</li><li>We know the n= ature of the speech in such a conversation, from a low-context vs. high-con= text perspective.</li><li>The talkers are within their critical distance (i= .e., ignore reverberation of talkers’ speech).</li><li>We know the lo= ng-term spectral shape of speech at the eardrums.</li><li>We know the long-= term spectral shape of noise at the eardrums.</li><li>The noise is stationa= ry.</li><li>The noise is diffuse.</li><li>The diffuse noise in combination = with diotic direct-wave target speech generates the equivalent of an “= ;internal” wideband binaural SNR improvement of ~1.0 dB over monaural= listening.</li><li>Overall level is not too high (i.e., no “roll-ove= r” effect for intelligibility has occurred).</li><li>We’re not = including lip reading.</li></ol> <div><font face=3D"Times New Roman"> </font></div> <div>These assumptions allow one to, first, determine the AI needed to achi= eve the assumed target %C given the assumed type of speech, and, second, de= termine the SNR required with the assumed spectral shapes to obtain that AI= .</div> <div> </div> <div>For instance, assuming talkers want 95%C with “unfamiliar senten= ces” then using the polynomial fits from Sherbecoe and Studebaker (JA= SA 1990) of the ANSI S3.5-1969 transfer functions between AI and %C, our ta= lkers would need an AI of 0.45. If we assume the noise and speech have the same long-term spectral shape then the SNR ca= n be determined from (SNR+12)/30=3D0.45 (staying with the 1969 AI metho= d), yielding SNR =3D 1.5 dB. Subtracting the binaural SNR improvement= yields 0.5 dB. </div> <div><font face=3D"Times New Roman"> </font></div> <div><font face=3D"Tahoma" size=3D"2"><span style=3D"font-size:10pt;"><b>Fr= om:</b> AUDITORY - Research in Auditory Perception [<a href=3D"mailto:AUDIT= ORY@xxxxxxxx">mailto:AUDITORY@xxxxxxxx</a>] <b>On Behalf Of <= /b>Andy Sabin<br> <b>Sent:</b> Wednesday, January 22, 2014 9:53 AM<br> <b>To:</b> AUDITORY@xxxxxxxx<br> <b>Subject:</b> Reference for typical SNRs is public spaces</span></font></= div> <div><font face=3D"Times New Roman"> </font></div> <div><font face=3D"Times New Roman">Hi List, </font></div> <div><font face=3D"Times New Roman"> </font></div> <div><font face=3D"Times New Roman">Can anyone point me to a reference show= ing SNRs that are typically observed in public spaces (e.g., restaurants, b= ars ...etc)? I can find this info for overall SPL, but am having a hard tim= e finding it for SNR. </font></div> <div><font face=3D"Times New Roman"> </font></div> <div><font face=3D"Times New Roman">Thanks</font></div> <div><font face=3D"Times New Roman">Andy Sabin</font></div> <div><font face=3D"Times New Roman"> </font></div> </span></font> </body> </html> --_000_8B7E4506479CA84B8A00E10CF01B1143D0FBF33Cep2pexmbs2mssta_--