Re: streams and groups ("John K. Bates" )

Subject: Re: streams and groups
From:    "John K. Bates"  <jkbates(at)COMPUTER.NET>
Date:    Tue, 15 May 2001 21:41:47 -0400

Dear Bruno, Al, and List, Thanks for your responses on my applause problem. However, subsequent definitions seem to be getting as amorphous as the applause. Perhaps this is because we might be thinking deeper into the structures of what constitute spatially-distributed impulsive events. (No lisp intended.) For example: >>In any case, further parsing of my applause >>would reveal that every individual clap is itself composed of streams >>and/or groups of smaller identifiable transient events. > >That seems unlikely to me: Each clap is just a single transient. >However, there are more complex auditory events that can indeed be >decomposed into smaller elements that are perceptually grouped >together. Yes. We could regard the single clap transient, which is not just a simple impulse function but is a complex structure of damped resonances and reverberations, as a stream of intermixed groups of events. The likelihood of parsing this stream successfully depends on how you approach the problem. We know it can be done because the ear has found a good way to do it. >I have never been in this situation (there is usually total silence >after I finish a talk), but I would say the mixture of sources >contains a large number of potential streams that the listener's >perceptual system is unable to segregate. The sources and streams are >physically there but difficult to detect. By placing a microphone a >few inches from the hands of any individual clapper, the stream of >his/her claps can be isolated and made perceptible. I was assuming an impersonal interest in the composition of the applause. More to the point, there seems to be a certain level at which our auditory systems decide that the number of clappers (or voices or raindrops) becomes a single applauding crowd...too many to segregate into individual sources or groups. (Apparently you and I have not reached that level.) Has anyone found that limit? Most such projects now seem, at best, to regard more than two sources as a crowd. The problem is significant in devising the algorithm for a real CASA system. - John Bates

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