Parametric Tuning of Extended Reverberation Algorithm Using Neural Networks

Student thesis: Master thesis (including HD thesis)

  • Søren Vøgg Krabbe Lyster
4. Term, Sound and Music Computing (Master Programme)
This thesis sets out to implement an extended feedback delay network with a comprehensive set of parameters that can be estimated by a proposed neural network. The goal of the neural network is to use audio differentiation to tune the parameters of the feedback delay network, to allow it to emulate other reverberators. The feedback delay network is implemented as a VST3 audio plugin and embedded in the neural network via a proposed audio processing model. Qualitative evaluation is done on the performance of the plugin and the neural network, and a perceptual listening test is done to evaluate the subjective quality of the reverberated signals created by the estimated reverberation parameters.
Publication date25 May 2022
Number of pages66
ID: 471369448