doug@xdos.UUCP (Doug Merritt) (02/03/90)
In article <632@xdos.UUCP> doug@xdos.UUCP (Doug Merritt) writes: >I'm interested in finding a discussion of the waveforms/spectral signatures >of various different kinds of sound effects, similar to the usual Ok, I got lots of interest in summarizing what I found out, so here it is: everyone wants such a thing, but no one knows of such a thing. So I've started writing some analysis software (I'm up to *here* in Digital Signal Processing books). I'm going to start by comparing the spectral signatures of the small number of samples I've already got, and categorizing them by their psychoacoustics as interpreted by my not-very-trained ear. I'm also going to do some automatic pattern matching to help find similarities that my ear and eye might miss (envelopes, spectral widths & ratios, purity/noise content, whatever else comes to mind). So, two more questions: does anyone know of reference material on any existing systems for doing the classification, or should I make up my own? And: I've never used synthesizers, but I know that somewhere out there are a huge number of samples. Can anyone give me any leads on how to get various samples (hopefully light on the acoustic instrument category), in the public domain to keep it legal, on some kind of common computer media? I could handle diskettes for IBM PC's, Mac's, Amiga's, 9 track tape or Sun cartridge tape. But I would not know what to do with some special purpose media designed only to work with a synthesizer. Ideas? I'm probably going to do a pamphlet or even a sourcebook on what I eventually come up with, since there seems to be a lot of interest. Thanks, Doug -- Doug Merritt {pyramid,apple}!xdos!doug Member, Crusaders for a Better Tomorrow Professional Wildeyed Visionary
EHARNDEN@AUVM.BITNET (Eric Harnden) (02/03/90)
i don't know of any source books as requested, but one thing does occur to me to note... while there may be a source of 'snapshot' analyses of the kinds of sounds you mention (although i kind of doubt it), the most critical thing about sounds of that sort are their dynamics, which are not very well (if at all) characterized in the literature. the other thing is that different categories of sounds are going to be approachable by vastly different analysis/ synthesis techniques. for example, to do slams and explosions i would probably combine functional modules for noise and damped sine generation in varying degrees, and pass their output through a digital filter with broadly tuned resonant formant regions. in such a case, spectral analysis per se would be of limited use, since the target timbres are 1)so dynamics-dependent 2)essentially completely inharmonically structured. on the other hand, bird whistles would be another thing entirely. see what i mean? so in your position (which, actually, i am... i'm pursuing a related project myself), i would start by providing myself with a generalized set of synthesis tools, developing an ear for sound deconstruction, and playing mix and match.