Combining a fuzzy logic engine and a neural network to develop an embedded audio synthesizer

The goal of this research is to integrate in a single audio synthesizer both instrument identity and player ability. A soft computing approach to match audio synthesis is then proposed. Combining soft decision capability of fuzzy logic and generalization ability of neural networks, a very powerful audio syn-thesizer can be modeled. In the proposed model, a fuzzy logic engine is tuned by means of fuzzy rules to generate controls to be applied to a neural network trained for audio synthesis. An audio control, a symbolic score or a text is fuz-zyfied to feed the fuzzy engine. This evaluates a set of soft control outputs to be inputted to the neural network that outputs a sequence of audio frames. A post-processor control and chain the audio frames for final audio synthesis.