High-Accuracy Spectral Modelling of Sounds

Currently, recorded sound samples dominate sound post-production for media. However, these are static and cannot be manipulated in the ways that sounds generated by synthesis engines can be. Such samples are inflexible and limited, and their repetitive use results in unnatural and dull sound scenes. Methods for modelling and synthesis of sounds have existed for many decades but can lack the realism of sampled sounds. This project takes work done in MPPAG on high-accuracy modelling of audio and deploys it in tools for audio post-production for multimedia applications. The project is in collaboration with RPPtv and is supported by an Innovate UK Knowledge Transfer Partnership grant.

KTP Project, lead academic: Dr Jez Wells

Knowledge transfer/research associate: Mr Ken Brown

The KTP project finished in February 2020 and was rated 'good' by Innovate UK.

Work continued in a six-month EPSRC Impact Accelerator Account project which specifically looked at offset estimation and frequency-domain synthesis. This project finished in October 2020. Libraries are available, under a GNU public licence (FHAMAS).