The purpose of this investigation is to find out if a large range of timbres can be learnt on signal level and then recreated generatively. Recorded sounds are often preferred over synthetic sounds by composers or sound designers because of their inimatable complexity. The most accurate way of recreating such complexity computationally is by constructing a physical model of the desired sound object(s). However, a physical model for one sound object is unlikely to be usable for an entirely different sound object. The more complex the desired sound is the more expensive and limiting the physical model will be also. This problem can be avoided if timbre can indeed be learnt and recreated.

For the learning process to be successful timbre should not simply be discernable by the system (MFCCs can already be used as a representation of timbre). Exact information about the "micro-level composition" (Döbereiner, 2011, p. 29) of the sound should be learnt for the purpose of it's subsequent reproduction...

I have launched my soundcloud and I currently have one track. For this track I have restricted myself to using my own software, the GA Gendy, in order to generate my source sounds. I have not used any recorded samples. I have been featured on the blog of Deiru Black who I met in brighton recently. The track "Drench" also gained some airtime on Guy Andrews' radio show The Drum Clinic.

I used SuperCollider to create my own artificial neural network class and trained it to stochastically alter waveforms and improvise upon given rhythms. The neural network is trained by back propagation. All of the synthesis generated is done at sample level with an idiosyncratic breakpoint synthesis algorithm which is periodically altered by the neutal net. The harmony, melody and tonality is organised and controlled at runtime by nested Markov models. The markov model also acts as an array, each item corresponding to the state of the Markov model (hence why Markov models can be nested).

The unifying theme of my project is my attempt to build as much source material as possible from the ground up. All of the synthesis is created by the... Click read more to hear an example.

GA Gendy is my research based final year project for my Music Informatics undergraduate degree. It is a stand-alone synthesizer program built in java based upon Xenakis' generative dynamic stochastic synthesis algorithm ("Gendyn") originally written in BASIC. The technique appears in compositions GENDY3 (1991) and S.709 (1994) (Luque, 2006, p. 5). I have then used an unsupervised genetic algorithm (GA) training procedure to try and train Gendyn's parameters to produce output that timbrally matches a supplied audio clip.

The program is written in java and has a fully functional graphical user interface, including a full graphical envelope system. Gendyn partly relies upon use of probability distributions and I have introduced...

I am Jonathan Young and I currently work as a programmer at PaperSeven in Hove. I am a University of Sussex Music Informatics graduate.

Languages:
Java, C sharp, Objective-C, PHP, MySQL, CSS & HTML, XML, AJAX, Javascript, Processing, SuperCollider.

Areas of Expertise:
App development (iOS and Android), OOP, Games Development in Unity, server side web computing and web design, machine learning & listening, digital audio processing.

This website serves as an online portfolio as well as providing updates for any projects I undertake.