Skip to main content
School of Electronic Engineering and Computer Science

Profile

Project title: Neural Audio Synthesis with Expressiveness Control 

C4DM theme affiliation: Audio Engineering, Music Informatics, Sound Synthesis. 

Abstract: This project will be focused on a prospective methodology for Neural Audio Synthesis, where the objective is to leverage prior knowledge and understanding about audio expression to build a neural audio synthesis framework with high controllability on expressiveness. More specifically, this project aims for a neural audio synthesizer that achieves the control over audio expression from fine-grained to coarse levels of information. The expected contributions include:  

  • introducing controllability over the non-harmonic acoustic components within the Differential Digital Signal Processing (DDSP) framework. 
  • leveraging extracted audio representations for efficient and robust style transfer. 
  • addressing the disentanglement of expression with respect to different time scales. 

Research

Back to top