Intended for developers :
The most important class (or file) in this project is the brain class. It defines basically a neural network with inputs, outputs and a function to make it process data.
The inputs are float, preferably beetween 0 and 1 (although there is an auto calibration system to make unspecified values to match [0,1]) The outputs are always float beetween 0 and 1.
test.c is a good and simple example of how it works. Take a look at this file, it's really simple !
From now on, it should be easy to generate weird sounds :
1st technique : Generating samples (I don't know if it oscillates enough to produce sound)
You have to create one brain with 1 or more inputs and 1 output (and as many neurons as you want...)
You write what you want in inputs (random float values preferably between 0 and 1, sinus waves, a sound sample, whatever).
You process one step
You read the output. It's a float value beetween 0 and 1. You convert it into one sample (basically by multiplying it by the max sample : 65536 for 16 bit samples), and you write that value to a well formated sound file.
Repeat as long as you want !
2nd technique : Generating frequences (It oscillates by definition).
You have to create one brain with 1 or more inputs and probably 3 outputs : one for the frequency, one for the length and maybe one other for the volume.
There are other sound synthesis techniques, feel free to implement them ! You can even try to make realtime sound synthesis, it would be really great :)
ikari <ikari@ifrance.com>