Sun Aug 14 16:13:33 CEST 2011
Audio DSP is not Image DSP
I came to a realization today that audio DSP and image DSP are
fundamentally different. The reason is causality - the directionality
of the time variable.
What this means is that for Audio, the principle abstraction is the
difference equation, a recursive relation between current and past
When this recursive model is not imposed or "strongly pulled" by the
problem domain, other "symmetric" models are possible that allow for
different divide and conquer approaches.
With "strongly pulled" I mean that a lot of algorithms have more bang
for the buck if recursion is allowed: reusing past solutions in
current ones. What this means practically for my DSP language is that
I probably don't get far without acknowledging the direction of time,
and using the state monad as a basic model.