Spatial Audio Coding and Machine Learning

Abstract

Spatial audio encoding plays a fundamental role in the ultra-high-definition TV (UHDTV) and the latest generation of television broadcasting, as well as other technological devices by providing a three-dimensional (3D) audio content to consumers. In this article, the fundamental concepts of the spatial audio coding including its techniques, standards, and applications are exhibited. The object-based audio reproduction system will be presented and compared to the traditional channel-based system in order to offer a good understanding of this system to the users and to give them more flexibility in their preferred audio composition. Moreover, the MPEG standard for encoding multi-channel audio signals will be exposed. Machine learning (ML) methods and their applications in acoustics and spatial audio scenes will then be offered. Ultimately, further research directions will be illustrated and discussed.