Neural networks for a simpler control of synthesis algorithm of musical tones and for their compression

Presents a hybrid system: a neural network for data compression and a generic algorithm for sound synthesis. The goal is the reduction of the number of parameters associated with a sound. Because of the growth of electronic technology, actual synthesisers are very complex systems in which the user cannot move without difficulty, and expensive loss of time. The basic ideas are reduction of the parameters made by the neural network (NN), and the freedom to choose any algorithm for sound synthesis. For example, using a machine based on FM synthesis, one can change a subset of the proper FM parameters, while in a sample-based keyboard one can change the gain value, some envelope parameters, etc. Such a system is very easy to use, flexible, and requires only a little more hardware for its realisation in an electronic synthesiser. The major work is for the developer, but this is made only once. The user just moves some controllers, i.e. some sliders; the neural network computes the output which is the set of the parameters for the synthesiser