Re: SERAF - Synaptic Energy Redistribution Audio Filter

stephan.bernsee@web.de  (Stephan M. Bernsee) wrote in message
news:<38ab652c.0405031006.579080d8@posting.google.com>
>
>Sounds interesting, although I see no real connection to the <Neuron>
>synthesizer from what I've read here. To be honest, I don't quite
>understand how this SERAF is supposed to work anyway - do you have
>more information? The bits I was able to dig out in this thread don't
>make much sense to me.
>

Thanks for your feedback.

http://groups.google.com/groups?q=author:stpnrrs%40aol.com
and
http://groups.google.com/groups?q=author:apollonicon%40aol.com
give Google's archive of the relevant threads (sorry about the slight mess, I
was experimenting with Google's newsgroup seach engine) and the the original
message at the start of the thread is reprinted below. 

The content of the threads covers most of what I wanted to say, without being
too definite about precisely how a SERAF might work (as I don't yet know!).  If
there are any paragraphs or ideas that you would like me to explain more fully,
I'll be more than happy to clarify them and give any further information you
may need..

***

Help needed for an audio project
Psychoacoustics project

I need some help to check out a theoretical model of hearing based on the
ear/brain's synaptic responses to sound.  Anyone familiar with audio software
may be able to help out by suggesting how to perform some experiments, or by
pointing in the direction of similar research that has already been done.

In its simplest form the theory is as follows:  An individual cilium in the ear
responds to variations in air pressure by causing a synapse to fire whenever
the cilium has been stimulated for a duration dependent on the amplitude of the
pressure; conceptually, any amplitude/duration pair has an associated synapse
which fires whenever that pressure amplitude occurs for that duration within a
sound.  The brain matches the patterns of synaptic firings caused by a sound
against remembered firing patterns when deciding on what a sound 'is'. 

Using computers, it may be possible to redistribute the amplitude/duration
information within a natural sound (such as a human voice saying the word
'three'), so that the brain matches remembered synaptic firing patterns in an
abnormal way, thereby triggering lower-level functioning of the brain and
evoking unusual sound qualities.  

For example, one could try moving all the waveform peaks to the left or right
so that the resultant waveform always has leading or trailing vertical edges at
the peaks.  Such sounds don't occur in nature, so one might hope that the brain
would identify something unusually interesting about the waveform (the synaptic
firings might be expected to occur in bunches at the peaks rather than being
more spread out), but as no-one has ever reported any successful results from
this fairly obvious experiment, the effect must be too subtle or random to be
useful, and something cleverer is required. 

The human ear/brain cannot distinguish between the onsets of sounds separated
by less than about a twenty-fifth of a second, so progress might be possible if
one allows amplitude/duration pairs to be moved by up to a twenty-fifth of a
second from their original position.  Starting with the digitised waveform,
generate lists of durations for each measured pressure amplitude (65536 lists
for 16-bit sound samples), then reorganise the pressure durations where
possible so that most of the energy occurs towards the start (or end) of each
25th of a second.  Depending on the chosen algorithm, the resultant waveform
can have leading or trailing vertical edges and the occasional micro-silence,
and these together with the (inaudible) 25 Hz pulse effect can be expected to
trigger extra brain activity because of the unusual sound structure (which
retains much of the original synaptic firing information), evoking the
hoped-for unusual sound quality.  There are other well-known methods of
subliminally stimulating the brain, such as forwards and backwards audio
masking techniques which train the brain to be more or less receptive to
hearing specific sounds, and these may well have an analogy at the synaptic
level which can also contribute to an evoked sound quality.  
The 'holy grail' of the project is to generate a waveform which has a bright,
open, honest sound when played forwards, and a warm, dark, secretive sound when
played in reverse.  A subsidiary aim is to develop a digital filter which
generates symmetric maxima in an audio waveform (instead of allowing leading or
trailing synaptic firing edges to produce special effects) so that an encoder
can easily compress the waveform data to about half its original size, without
loss of clarity.  These ambitious aims are unlikely to be completely a mirage,
given the brain's ability to dream and imagine, of its own volition, very
unlikely sounds which are presumably constructed only from remembered synaptic
firing patterns corresponding to previous auditory experiences. 

As presented above, the theory may be too over-simplified to be practical, but
if anyone gets a result from applying the ideas or needs more information,
please let me know!

***