Re: [AUDITORY] Localizing smoke detectors - why is it so hard? (John Culling )

Subject: Re: [AUDITORY] Localizing smoke detectors - why is it so hard?
From:    John Culling  <cullingj@xxxxxxxx>
Date:    Tue, 25 Jun 2013 13:38:09 +0100

It's probably the combination of a sparse frequency spectrum and room reverb. The tone is at about 3 kHz with limited harmonic structure, so there may be a recognisable ILD, but you can't tell = elevation=20 or front from back. The beep is brief, so you can't use head movement to disambiguate front and back. Then, the reverb will reduce the ILD = and=20 make it change from point to point in the room. The precedence effect should compensate for that to some extent, but the resulting perceived=20 azimuth is less stable for a tone than for a broadband signal. See, for instance (fig 1.).... _Blauert08/blauertbraasch2005.pdf If only it would produce a click train instead. John. =20 >>> "Jennifer M. Groh" <jmgroh@xxxxxxxx> 25 June 2013 09:46 >>> Dear List, I am writing a book for a general audience on how the brain processes=20 spatial information ("Making Space"). The chapter on hearing covers = many=20 topics in sound localization, but there is one that I'm currently still=20 quite puzzled about: why it is so hard to localize a smoke detector = when=20 its battery starts to fail? Here is what I have considered so far: - To my ear, the chirp sounds high frequency enough that ILD cues = should=20 be reasonably large. - At the same time, it seems to have a broad enough bandwidth, and in=20 any case it has onset-and-offset cues, that ITD cues should be usable. - A possibility is that the chirp is too brief, and that limits dynamic=20 feedback, i.e. changes in ITD and ILD as the head turns during a sound.=20 However, in my laboratory we have obtained excellent sound=20 localization performance in head-restrained monkeys and human subjects=20 localizing sounds that are briefer than the reaction time to make an=20 orienting movement. - An additional possibility is that we have too little experience with=20 such sounds to have assembled a mental template of the spectrum at the=20 source, so that spectral cues are of less use than is normally the case. I'm leaning towards a combination of the last two factors, which=20 together would render the cone of confusion unresolved for these stimuli. Thoughts? Best wishes, --Jennifer Groh --=20 Jennifer M. Groh, Ph.D. Professor Department of Psychology and Neuroscience Department of Neurobiology Center for Cognitive Neuroscience B203 LSRC, Box 90999 Durham, NC 27708 919-681-6536

This message came from the mail archive
maintained by:
DAn Ellis <>
Electrical Engineering Dept., Columbia University