Virtual Sound

A number of small bugs were fixed. A
VibratingString class was added for simulating
string and wind instruments. A small example is
given in VSound.c.

The stale file test_decoding.c (prevented compilation) and useless file accounting.c have been removed. Work has begun on a simple multiple-text search/clustering algorithm that will utilise the Graph class and belief propagation. This has nothing to do with sound, but it is an interesting application of probabilistic estimation. It will probably make sense to split the probabilistic and machine learning library in the future from the sound-related code.

More work was done on the statistical learning
feature. A few more examples were added. A class
for doing reinforcement learning was added.

Some low-to-mid abstraction layers were added.
These include the Single Input Multiple Output
Buffer and the Window class. SIMO mediates IO
between filters. Window is an abstraction allowing
filters to use erm, filtering windows such as a
Hamming window when reading data. This code has
not been really tested.

This release adds an artificial neural network with hidden control and predictive classes for modeling/prediction/classification of sounds. Many programs that use the new code have been added. It now depends on libsndfile-1.0.0.