Noise Reduction: From Capacitors to Convolutional Neural Networks

Noise Reduction technology kick-started Dolby Laboratories 55 years ago, and in 2020, continues to be a focus, albeit in completely new contexts and applications. Despite advances in capture technology and the shift in recording mediums, the proliferation of content creators and capture devices has seen an ever increasing need for solutions to reduce noise in audio. In this presentation, Nick Engel dives into noise reduction techniques – from analog processing through to today's machine learning approaches. He discusses his own experiences as an engineer working on the digital implementation of Dolby B Noise Reduction, through to more recent experiences for helping to reduce noise in audio recordings at scale with Dolby.io. This talk explores how the types of noise and audio capture and recording methods have progressed, as well as approaches for reducing noise, from analog, to digital signal processing and through to the state of the art of deep learning techniques.