[sdiy] Vocoder & neural network...

[Chris Stecker]

Sure, this is quite possible, and despite what some replies have suggested, 
you could do a lot of interesting things with this approach. Think of the NN 
as reducing the dimensionality of the data from N filters to M (<N) control 
parameters. 

The folks at CNMAT have used a number of schemes to control some really 
amazing additive synthesis technology, including training NNs to generate 
additive parameters based on analysis of real signals (instruments, etc.) 
This is basically what you describe. Check them out at 
http://www.cnmat.berkeley.edu. The technology evolved through several names: 
FAR (Fourier Analysis-Resynthesis), CAST (CNMAT Additive Synthesis Tech), 
etc, and has something to do with OSC (Open Sound Control) and SDIF (Sound 
Description Interchange Format). Just poke around the website a bit. The 
results were very interesting.

-Chris


-Chris

On Tuesday 11 March 2003 03:49 pm, jbv wrote:
> Hi all,
>
> I'm presently reading some litterature about
> neural networks, and was wondering about
> the possibilty to hook a neural network after
> the analyzer section (BPF bank + rectifiers)
> of a vocoder...
> The idea is to use the network for pattern
> recognition, collect the data and then drive
> a synthesizer section made of a limited bumber
> of BPF used has formants...
> The idea might also work with an FFT ouput...
>
> Has this been tried already ?
>
> JB