Re: Huggins pitch with speech? (John Culling )


Subject: Re: Huggins pitch with speech?
From:    John Culling  <cullingj@xxxxxxxx>
Date:    Fri, 28 Sep 2012 14:44:34 +0100
List-Archive:<http://lists.mcgill.ca/scripts/wa.exe?LIST=AUDITORY>

Dear Michael My recollection of Schroeder's patent is that it is not quite as you may imagine. However, I attempted to generate speech purely through binaural interaction some years back and published the results in JASA http://dx.doi.org/10.1121/1.2140806 Intelligibility was less than 10% for IEEE sentences. The best transcript from the entire study was "we all have a new day’s decision" for "we now have a new base for shipping" There are probably two reasons why we did not get very good intelligibility with purely binaural information. First, binaural unmasking is only effective at low frequencies (<1500 Hz). Consequently the stimuli all sounded low-pass filtered or like someone mumbling indistinctly. Second, our ideas at that time of what constituted binaural information were probably incorrect. The stimuli were designed to match the spectrotemporal variation in interaural correlation induced by the presence of the speech. Later experiments (also in JASA) suggest to me that the binaural system in fact works from the residue of interaural cancellation. http://dx.doi.org/10.1121/1.2785035 One might think, "OK let's manipulate that instead", but it is somewhat more difficult to independently control. One either has to allow the overal stimulus to retain some monaural spectrotemporal modulation (unlike a Huggins pitch) or the "noise" has to be allowed to modulate out-of-phase with the speech, which is rather unecological, but would at least give an unambiguous result - I haven't tried it, but the first problem would still need to be addressed. I may still have the stimuli somewhere, but be prepared for disappointment! John. >>> Michael Buschermöhle <m.buschermoehle@xxxxxxxx> 28 September 2012 13:31:03 >>> Dear list, I just recently stumbled upon the following book chapter by Manfred Schroeder: http://books.google.de/books?id=cg7x5_-vB-AC&pg=PA190&lpg=PA190&dq=manfred+schr%C3%B6der+huggins+pitch&source=bl&ots=tjbeher3kK&sig=TDtA5RUN1VLztWX55rKwKy-Hy_Q&hl=de&sa=X&ei=n1xkUOXhDsPOtAaeo4CoBw&ved=0CDEQ6AEwAA#v=onepage&q=manfred%20schr%C3%B6der%20huggins%20pitch&f=false In that chapter (Chapter 10.6) he explains that an effect similar to the Huggins pitch can also be evoked with speech (i.e. you have two unintelligible mono noise signals that reveal intelligible speech if you listen binaurally). As far as I understand this should be different (albeit related) to the BMLD effect. Manfred Schroeder cites the following patent which is supposed to be able to generate this effect: http://www.ptodirect.com/Patents/3328526 Does anyone know a sound sample or even a piece of source code that demonstrates this effect? I think it would be a wonderful effect to demonstrate the fascinating binaural processing that goes on in our brains. Thanks in advance for any reply! With kind regards, Michael. -- Dr. Michael Buschermöhle HörTech gGmbH c/o Haus des Hörens Marie-Curie-Straße 2 D-26129 Oldenburg Fon 0441/2172-210 Fax 0441/2172-250 eMail: M.Buschermoehle@xxxxxxxx HRB 5035, AG Oldenburg, Geschäftsführer: Stephan Albani


This message came from the mail archive
/var/www/postings/2012/
maintained by:
DAn Ellis <dpwe@ee.columbia.edu>
Electrical Engineering Dept., Columbia University