[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[no subject]

Dear Ramdas Kumaresan,

On Wed, 07 Mar 2001 you wrote:
>I agree with the general tone of your postings. I have expressed similar
>sentiments in my original posting (analog delay line vs 'continuously
>distributed memory'). However,  IMO, 'the signal is not simply
>transformed at all' is not true. After all the cochlea does have
>resonant structures and hence does some form of filtering  (which is
>convolution with damped sinewaves, i.e. a  form of weighted Fourier
>transform.)  I would like to fall into your third category.

This was my third category:

>>Hopefully, at least a few people will be able and willing to attack the
>>problems the other way round, that is, based on what is already
>>functionally understood, fairly regardless of the traditional approach and
>>the high level of sophistication in mathematics.

Well, I feel we are close. When I wrote 'not simply transformed at all', I
tried to object against the illusive hope that something like the Fourier
transform will be suited to actually mimic the function of cochlea. Look at
David McGrath's solution (US Pat. 5,502,747) to the comparable easy problem
of building a FIR-HRTF filter. It consists of a first time-domain
low-latency filter and a number of FFT fast-convolution filters. Dolby
headphones are based on this rather cochlea-lookalike structure. About
simultaneously (in 1995) William Gardner of MIT proposed in JAES nearly the
same hybrid structure as to compensate for the reverberation delay. Neither
digital time-domain filters nor fast-convolution filters are optimal in
real-time operation. The fast-convolution filters are not fast enough. You
are certainly also aware of acausality with frequency-domain filtering.
Narrowing my mind to the horizon set by tools like digital signal
processing and mathematical analysis, I would not be able for unbiased
understanding of what is going on. That's why I am pointing my finger
squarely to what I consider a fundamental difference between the presumed
function of the inner ear including pre-selection, amplification and
adjustment over a huge dynamic range, preparation for robustness and
adaptation to massively parallel processing on one side, and any mutilated
feet of Cinderella's sisters on the other side. I am convinced: Any
straightforward blind mathematical approach is doomed to fail. Any
convenient simple transform from time domain into any mathematical
construct will not adequately fit the function of hearing. As soon as we
were or will be able to largely understand auditory function, we tried or
will try inventing something that fits roughly or more reasonably,