Zoesch, Bob Katz, Steve Parker, et al,
The discussion came up with Bob Cain in respect to analyzing how we hear. We do not really hear in the Fourier Transform method that is the very basic model of hearing. We have a non-linear system that decomposes frequencies much like a Fourier Transform, but not as linearly as FT would imply. For this reason we can modify our model in one of two obvious ways - we can add all sorts of qualifiers to the FT to try to incorporate Fletcher Munson and other non-linearities in the hearing system, or we can use another tool that might be a modified FT that breaks the frequency spectrum, not into sine waves, but into some other waveform such that a simple analysis yields the complex results that are more akin to human hearing.
Again, we can either take a transform tool of simple waveforms (sine waves) and make it very complex with all kinds of qualifiers, or we can use a simple transform tool of a complex waveform (wavelets). Either way we get the same results, though through different means. For various reasons one tool might be more appropriate at one time and another tool might be appropriate at another time.
Just for an example, we have been discussing on the pro-audio mailing list about this window of analysis in the ear's hearing that affects what we hear in special situations. Creating a FT based model of the ear would get very complex if we add this characteristic in. Using a wavelet based model may give a more accurate result with a simple analysis - so long as the wavelet model chosen was accurate for the situation. There are many other situations wherein the ear is non-linear and simple FT analysis doesn't really represent the ear's hearing well at all. On the other hand, the vast majority of our hearing IS fairly Fourier based, so it is still an effective tool for a lot of purposes when discussing an auditory model.
I like Bob Cain's idea of looking at human hearing from a wavelet model to perhaps uncover several areas of suspicion, including Bob Katz's theories about the benefit of more gentle filters. This issue clearly gets lost on basic Fourier analysis models of the ear but may become more obvious with a good wavelet model.
I have been following the other threads for some time and never saw anyone explain where Bob Cain's approach to the problem was "wrong." I certainly don't see any evidence that this idea he puts forth is worthy of being called a "crackpot." It is far more valid from my assessment than many of the other theories about audibility beyond the "audible range."
Zoesch and Steve Parker - this sounds like something up your alley. Your thoughts? Should we start a new thread?
Nika.