Method for processing 3D audio

Abstract

Disclosed is a method and system for upmixing a first ambisonic signal to a second, higher-order ambisonic signal. It may utilise a trained machine learning model, such as a variational autoencoder, and may be used within a video game system. The upmixing process may be based on the type of audio signal, such as ambience, dialogue, music, or player sound effects. Different types of audio may be upmixed to ambisonic signals of different orders. The first ambisonic signal may be stored in memory whilst the second ambisonic signal may be decoded for playback. Also disclosed is a method of training a machine learning model to upmix a first ambisonic signal to a second ambisonic signal, wherein a first ambisonic representation of a sound is used as a training input and a second ambisonic representation is used as the intended output.

Images